Claude Skill

improve-content

Use when rewriting or refreshing an existing page that's underperforming. The agent fetches the URL, analyzes the current content, researches the SERP, and rewrites using the full anti-AI-slop ruleset — no data exports needed.

LLM Mart · 0 points · 7 views 0 listing impressions 0 install-command copies
Virus-scanned Reviewed automatically before listing.

Full trust report

Download inhouseseo-superseo-skills-skills_improve-content-9cf22cc.zip · 78 KB
Part of inhouseseo/superseo-skills — 11 skills

Install

skills CLI npx skills add https://github.com/inhouseseo/superseo-skills/tree/main/skills/improve-content
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install inhouseseo-superseo-skills@llmmart
Git git clone https://github.com/inhouseseo/superseo-skills.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole inhouseseo/superseo-skills collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Improve Content

Rewrites an existing page so it outranks competitors — better structure, deeper expertise, and a voice that actually sounds human. Uses the same anti-slop ruleset as write-content.

Input

URL of the page to improve. If the fetch fails, ask the user to paste the page content directly.

Role

You are a senior content writer and SEO specialist with 10+ years improving underperforming pages.

Step 1: Read the Current Content

Fetch the URL and read the full rendered page. Note title, meta, H1/H2s, word count, internal links, schema. Identify the apparent primary keyword from the content.

Step 2: Research the SERP

Google the primary keyword. Read the top 5 results. Note:

  • What formats are ranking
  • Content gaps where the current page falls short
  • Angles competitors take that this page doesn't
  • Any featured snippet / PAA opportunities

This is a lightweight check, not a full page audit. Just enough to know what's missing.

Step 3: Update Interview

Ask 2-3 quick questions adapted for content updates, one at a time:

  • "What's changed since you published this? New data, new experience, market shifts?"
  • "What results did this content get? Traffic, leads, feedback?"
  • "Knowing what you know now, what would you add or cut?"

If the user can't answer — autonomous run, or they skip — go straight to the rewrite and note where fresh first-party input would strengthen it.

Step 4: Rewrite

Produce the complete rewritten article in clean markdown. Apply ALL writing rules:

Voice

Practitioner to a peer. Take positions. Use "you" and "I/we." Specific numbers, names, dates — but only real ones, taken from the existing page, the SERP research, or the user's answers. Never invent a number, name, or citation; cut the claim instead. Contractions. Show thinking changing. Shift registers — after technical explanation, drop into casual aside. Uniform register = AI tell.

Rhythm

Vary sentence length dramatically (5 to 30+ words). Vary paragraph length. Fragments. Parenthetical asides. Break the topic-sentence-support pattern. Cover sections asymmetrically. No section summaries.

Show, don't state

Brief scenarios instead of flat claims. "You click a search result. Three seconds pass. Still loading. You hit back." narrates; "page speed affects UX" states.

Anti-slop

NEVER use: delve, landscape (metaphorical), testament, leverage, utilize, robust, seamless, furthermore, moreover, additionally, pivotal, multifaceted, harness, embark, navigate (metaphorical), showcase, streamline, paramount, culminate, spearhead, commence, endeavor, vibrant, innovative, comprehensive (as adjective).

NEVER use: "It's worth noting", "In today's [anything]", "Let's dive in", "In conclusion", "plays a crucial/vital role", "It goes without saying".

Avoid: rule-of-three groupings, synonym cycling, em-dash chains (max 1-2 per 1,000 words), binary contrasts, participial tack-ons.

SEO

Primary keyword in H1, first 100 words, 2-3 H2s. ~2% density, naturally distributed. 40-60 word direct answer after the most important H2 (featured snippet). Weave PAA questions as H2/H3 headings. 3-5 internal links per 1,000 words. Front-load value.

Content type

Detect from the existing page + SERP. Load references/content-types-overview.md for the decision table, then load the specific template from references/content-types/<type>.md for the full rewrite structure. 19 templates bundled: how-to, definition, comparison, listicle, pillar-page, faq-page, landing-page, service-page, case-study, statistics-page, news-article, glossary-page, alternatives-page, buying-guide, product-page, category-page, integration-page, location-page, programmatic-page.

  • How-to: 40-60 word quick answer, then numbered steps with "what goes wrong"
  • Comparison: verdict first, then detail
  • Definition: "[Term] is..." in the first sentence
  • Case study: result number first, then the story

Final checks

  1. "So what?" test per section — could anyone have written this for anyone? If yes, inject specific knowledge.
  2. Self-check: scan for blacklisted words, topic-sentence patterns, section summaries. Fix before delivering.

Output

Clean markdown. Title + rewritten article content. Nothing else.

Language

Write in the same language as the existing page content.

Bundled references

Load from references/ only when needed — don't preload the whole folder.

Content type templates (references/content-types/) — load one in Step 4:

  • Common: how-to.md, definition.md, comparison.md, listicle.md, pillar-page.md, faq-page.md, landing-page.md, service-page.md, case-study.md
  • Content and news: statistics-page.md, news-article.md, glossary-page.md
  • Commercial: alternatives-page.md, buying-guide.md, product-page.md, category-page.md, integration-page.md, location-page.md, programmatic-page.md
  • references/content-types-overview.md for the decision table across all 23 content types (load this FIRST if unsure which type to pick)

Rewriting technique modules (references/) — load when a rule calls for more depth:

  • anti-slop-ruleset.md — full tiered banned vocab + structural tells (load when the inline anti-slop block isn't catching everything in the current page)
  • voice-injection-playbook.md — how to inject voice into a flat existing page without rewriting from scratch
  • content-freshness.md — decay signals and what to refresh vs. leave alone (the core of Step 3's update interview)
  • writing-pipeline.md — the research → rewrite → edit loop adapted for updates
  • quality-scoring.md — scoring rubric to run before delivering the rewrite
Files (superseo-skills)
  • references
    • content-types
      • alternatives-page.md 5.4 KB
        # Technique 28: "Alternatives to X" Pages
        
        ## What It Is
        Commercial investigation content listing and evaluating alternatives to a specific well-known product or service. Captures "[Product] alternatives", "[Product] vervangen", and "iets zoals [Product]" searches. Targets users dissatisfied with or priced out of a market leader.
        
        ## When to Use
        Commercial investigation intent -- user already knows a specific product but is looking for other options. Triggered by dissatisfaction, price concerns, feature gaps, or general curiosity. High conversion potential because the user is already in buying mode.
        
        ## Structure Template
        
        ```
        H1: "[X] Alternatieven: [N] Beste Opties in [Year]"
            (e.g., "Vattenfall alternatieven: 8 energieleveranciers vergeleken in 2026")
        
        H2: Why look for [X] alternatives?
            Top 3-5 reasons people switch (pricing, features, support, etc.)
            Validates the reader's reason for being on this page
        
        H2: Quick comparison table
            All alternatives vs the original product
            Key criteria: price, features, ease of use, rating
            Bold the winner per criterion
        
        H2: 1. [Alternative 1] -- Best for [specific use case]
            H3: Overview and how it compares to [X]
            H3: Key advantages over [X]
            H3: Drawbacks compared to [X]
            H3: Pricing
        
        H2: 2. [Alternative 2] -- Best for [specific angle]
            Same structure
        
        H2: 3-8. [Remaining alternatives]
            Same structure, can be shorter for lower-ranked options
        
        H2: [X] vs the competition: feature comparison
            Detailed feature-by-feature comparison table
            Objective criteria, not opinion
        
        H2: How to choose the right alternative
            Decision tree or criteria checklist
            "Choose [Alt 1] if..." / "Choose [Alt 2] if..."
        
        H2: FAQ
            "Is [Alt] cheaper than [X]?"
            "Can I migrate from [X] to [Alt]?"
            "What's the best free alternative to [X]?"
        
        H2: Our recommendation
            Clear verdict based on use case
        ```
        
        ## Word Count
        1,500-3,000 words (150-300 words per alternative plus intro/comparison sections)
        
        ## Schema Markup
        - **Primary:** ItemList (with ListItem entries)
        - **Secondary:** FAQ (for FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** List snippet (alternatives queries trigger numbered lists)
        - **Target:** The numbered list of alternatives with their "best for" labels
        - **Alternative:** Table snippet for the comparison table
        
        ## CTA Placement
        - After each alternative: "try [alternative]" or "compare prices" link
        - After comparison table: quiz or decision tool
        - End of article: recommendation with conversion CTA
        
        ## Internal Linking Strategy
        - Link to individual review articles for each alternative
        - Link to "[X] vs [Alternative]" comparison articles
        - Link to the review of the original product [X]
        - Link to buying guide for the category
        - Receive links from review pages and "best of" roundups
        
        ## Key Success Factors
        1. **Name the reasons for switching:** Validate why the reader is looking for alternatives
        2. **Direct comparison to [X]:** Each alternative must be explicitly compared to the original
        3. **Specific "best for" labels:** Do not just list alternatives -- explain who each one is best for
        4. **Include the original product:** Sometimes [X] is still the best choice -- acknowledging this builds trust
        5. **Migration guidance:** Address how to switch (data export, learning curve, contract terms)
        6. **Honest about trade-offs:** Every alternative has something it does worse than [X]
        
        ## Common Mistakes
        - Listing alternatives without comparing them to the original product
        - No clear recommendation or decision framework
        - Missing the "why switch" section (does not validate the reader's intent)
        - Not addressing migration or switching costs
        - Listing too many alternatives (5-8 is optimal; 15+ suggests no curation)
        - Biased toward one alternative without transparency
        - No pricing information (the #1 reason people look for alternatives)
        
        ## Anti-AI Focus
        Alternatives pages require genuine knowledge of both the original product and its competitors. To establish authenticity:
        
        - **Explain the specific pain points that drive switching.** "Vattenfall raised variable rates by 12% in Q1 2026, which is why we are seeing increased interest in alternatives" -- this kind of market-aware context is beyond AI's reach.
        - **Document your migration experience.** If you have switched from X to an alternative, describe the process: how long it took, what data transferred, what was lost, and what surprised you.
        - **Include real pricing comparisons with dates.** "Checked on March 15, 2026: Provider A charges X/month vs Provider B at Y/month for comparable plans." Dated pricing is a strong authenticity signal.
        - **Note what the alternatives do worse.** AI-generated alternatives pages tend to be uniformly positive about all options. Honest drawbacks ("Alternative 3 has better pricing but their mobile app is significantly worse than X's") signal real evaluation.
        - **Reference user community sentiment.** "On Tweakers, users report that Alternative 2's setup process takes about 30 minutes longer than X's" -- citing specific community feedback demonstrates research depth.
        
        ## Example Topics by Niche
        - Energy: "Vattenfall alternatieven: 8 energieleveranciers die goedkoper zijn"
        - Telecom: "Ziggo alternatieven: de beste internet providers zonder Ziggo"
        - SaaS: "Ahrefs alternatives: 7 SEO tools compared on features and price"
        - E-commerce: "Bol.com alternatieven voor verkopers: 6 marktplaatsen vergeleken"
        - Local services: "Werkspot alternatieven: 5 platforms om een vakman te vinden"
        
      • buying-guide.md 5.8 KB
        # Technique 30: Buying Guides
        
        ## What It Is
        Educational content that helps users understand what to look for when purchasing a product or service category. Focuses on criteria, considerations, and decision-making framework rather than specific product recommendations.
        
        ## When to Use
        Commercial investigation intent -- user is early in the purchase journey and needs to understand the category before evaluating specific options. Captures "where to buy [category]", "[category] kiezen", "what to look for in [category]" searches.
        
        ## Structure Template
        
        ```
        H1: "[Category] Kopen: Waar Moet Je Op Letten?"
            (e.g., "Zonnepanelen kopen: waar moet je op letten in 2026?")
        
        H2: Quick checklist
            5-7 key criteria as a bulleted list
            Optimized for featured snippet
        
        H2: [Criterion 1]: [What to evaluate]
            H3: Why it matters
            H3: What to look for
            H3: Red flags to avoid
            Specific numbers, ranges, or benchmarks
        
        H2: [Criterion 2]: [What to evaluate]
            Same structure
        
        H2: [Criterion 3-6]: [What to evaluate]
            Same structure
        
        H2: Budget guide
            Price ranges by quality tier (budget, mid-range, premium)
            Total cost of ownership, not just purchase price
            Hidden costs to watch for
        
        H2: Common mistakes when buying [category]
            5-7 specific pitfalls with how to avoid them
        
        H2: When to buy / Timing
            Seasonal pricing, sales periods, optimal timing
            Dutch-specific timing if relevant
        
        H2: Where to buy
            Channels compared: online vs offline, specialist vs generalist
            Dutch-specific recommendations
        
        H2: Decision checklist
            Printable/downloadable checklist of all criteria
            Interactive if possible
        
        H2: FAQ
            5-7 pre-purchase questions
        
        H2: Ready to choose?
            Links to "best of" roundups, comparison pages, specific reviews
        ```
        
        ## Word Count
        2,000-3,500 words (comprehensive enough to cover all buying criteria)
        
        ## Schema Markup
        - **Primary:** Article (with author, datePublished, dateModified)
        - **Secondary:** FAQ (for FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** List snippet for "what to look for" queries
        - **Target:** The quick checklist section at the top
        - **Alternative:** Table snippet for the budget guide with price ranges
        
        ## CTA Placement
        - After quick checklist: downloadable PDF checklist
        - After budget guide: comparison tool or calculator
        - End of article: links to "best of" roundup and specific product reviews
        
        ## Internal Linking Strategy
        - Link to "best of" roundups for specific recommendations
        - Link to individual product reviews for deep dives
        - Link to comparison articles for head-to-head evaluations
        - Link to definition articles for technical terms
        - Link to statistics pages for market data
        - Receive links from pillar pages and product/category pages
        
        ## Key Success Factors
        1. **Criteria-focused, not product-focused:** Teach how to evaluate, do not recommend specific products
        2. **Specific benchmarks:** "Look for at least 370Wp per panel" not "look for good performance"
        3. **Budget ranges with context:** Price alone means nothing without explaining what you get at each tier
        4. **Red flags section:** What to avoid is often more useful than what to look for
        5. **Total cost of ownership:** Include installation, maintenance, running costs -- not just purchase price
        6. **Decision checklist:** A practical tool the reader can use while shopping
        7. **Seasonal/timing advice:** When to buy for the best deal (Dutch market-specific)
        
        ## Common Mistakes
        - Recommending specific products (that is a roundup, not a buying guide)
        - Vague criteria without specific numbers or benchmarks
        - Only covering purchase price, not total cost of ownership
        - Missing the "common mistakes" section (one of the most valuable parts)
        - No budget tiers (readers need to know what is realistic for their budget)
        - Too focused on features, not enough on real-world impact
        - Not addressing the "where to buy" question
        
        ## Anti-AI Focus
        Buying guides require domain expertise to write well because the value lies in knowing what actually matters versus what sounds important on paper. To ensure your guide signals genuine expertise:
        
        - **Include benchmarks from real market research.** "In the Netherlands, a quality solar panel installation costs between 4,000 and 8,000 euros for a typical household as of Q1 2026" -- specific, current, localized numbers that AI would need to guess at.
        - **Explain WHY each criterion matters from experience.** "We recommend prioritizing inverter quality over panel wattage because in our experience, inverter failures account for 80% of system issues in the first 5 years." The reasoning behind the recommendation is where expertise shows.
        - **Share red flags you have personally encountered.** "Be wary of installers who quote without a roof inspection -- we have seen this lead to structural issues in at least 3 cases." Real cautionary examples are powerful authenticity signals.
        - **Include Dutch-specific buying context.** Mention relevant subsidies (ISDE), regulations, seasonal patterns, and local market dynamics. AI-generated buying guides tend to be geographically generic.
        - **Reference price movements over time.** "Panel prices dropped 15% between 2025 and 2026, so waiting for further drops is less likely to pay off than it was two years ago." This temporal market awareness demonstrates ongoing engagement with the category.
        - **Add the "questions to ask the seller" section.** Specific questions that expose quality differences ("Ask for the panel degradation rate after 25 years -- anything above 0.5% per year is below industry standard") demonstrate insider knowledge.
        
        ## Example Topics by Niche
        - Energy: "Zonnepanelen kopen: 8 criteria waar je op moet letten in 2026"
        - Telecom: "Internet abonnement kiezen: de complete koopwijzer"
        - SaaS: "How to choose a CRM: the complete buying guide for SMBs"
        - E-commerce: "Laptops kopen: waar moet je op letten? Koopgids 2026"
        - Local services: "Aannemer kiezen: 10 dingen waar je op moet letten voor je verbouwing"
        
      • case-study.md 5.1 KB
        # Technique 38: Case Study Pages
        
        ## What It Is
        Evidence-based content showcasing a specific client result, project outcome, or success story. The most powerful E-E-A-T content type because it demonstrates real Experience and Expertise with verifiable results. Captures "[service] results", "[industry] case study", and "[problem] oplossing" searches.
        
        ## When to Use
        Commercial investigation intent -- user wants proof that your service or product delivers results before committing. Use when you have specific, measurable client outcomes you can share. Essential for high-consideration purchases (B2B, services, expensive products).
        
        ## Structure Template
        
        ```
        H1: "[Client/Industry]: [Key Result in Numbers]"
            (e.g., "Energievergelijker.nl: 142% meer organisch verkeer in 6 maanden")
        
        H2: Overview / At a glance
            Client: [name or anonymized descriptor]
            Industry: [sector]
            Challenge: [1 sentence]
            Solution: [1 sentence]
            Results: [3-4 key metrics]
            Timeline: [duration]
        
        H2: The challenge
            What problem the client faced
            Context: company size, market, previous attempts
            Why it mattered (business impact of the problem)
        
        H2: Our approach / The solution
            H3: Phase 1: [Discovery/Analysis]
                What we did and why
            H3: Phase 2: [Implementation]
                Specific actions taken
            H3: Phase 3: [Optimization/Scaling]
                Refinements and scaling
        
        H2: The results
            Primary metric: headline number with context
            Secondary metrics: 3-5 supporting data points
            Visualizations: charts, before/after screenshots
            Timeline: how quickly results appeared
        
        H2: Key takeaways
            3-5 lessons that apply to similar businesses
            Actionable insights for the reader
        
        H2: Client testimonial (if available)
            Direct quote from the client
            Name, title, company (with permission)
        
        H2: Want similar results?
            CTA to relevant service page or consultation
        ```
        
        ## Word Count
        1,500-2,500 words (detailed enough to be credible, focused enough to maintain interest)
        
        ## Schema Markup
        - **Primary:** Article (with author, datePublished, about)
        - **Secondary:** Organization (for the client, if named)
        
        ## Featured Snippet Strategy
        - **Format:** Paragraph snippet for "[industry] case study" queries
        - **Target:** The overview/at-a-glance section with the key result
        - **Tip:** Include the primary metric in the first 50 words of the page
        
        ## CTA Placement
        - After overview (above fold): "Want similar results? Contact us"
        - After results section: primary conversion CTA (strongest proof point)
        - End of page: consultation or service page CTA
        - Sidebar: related case studies in the same industry
        
        ## Internal Linking Strategy
        - Link to the service page for the service described in the case study
        - Link to related case studies in the same industry or service
        - Link to methodology or how-to content explaining the approach
        - Link to pillar page for the topic area
        - Receive links from service pages, about page, and industry-specific content
        
        ## Anti-AI Focus
        Case studies are nearly impossible to AI-generate convincingly without real data. Real names (with permission), real numbers, and real timelines are essential. The specificity of a genuine case study -- exact percentages, named tools used, specific challenges encountered mid-project, and direct client quotes -- is what makes it credible. Vague or rounded numbers, generic industry descriptions, and unnamed clients are telltale signs of fabricated content. Invest the time to document real outcomes with real details.
        
        ## Key Success Factors
        1. **Lead with the number:** The headline result must be in the H1 and above the fold
        2. **Specific, verifiable metrics:** "142% increase in organic traffic" not "significant improvement"
        3. **Before and after:** Show the starting point and the end result with data
        4. **Process transparency:** Explain what you actually did -- not just that you did "great work"
        5. **Timeline included:** How long it took to achieve the results (sets realistic expectations)
        6. **Client permission:** Named clients with testimonials are vastly more credible than anonymous ones
        7. **Visual proof:** Charts, screenshots, before/after comparisons
        
        ## Common Mistakes
        - Vague results without specific numbers ("improved their rankings" means nothing)
        - No process description (readers cannot evaluate your expertise without knowing what you did)
        - Missing timeline (results without timeframe are meaningless)
        - No client quote or testimonial (misses the strongest trust signal)
        - Writing about yourself instead of the client's journey and results
        - Not linking to the relevant service page (missed conversion opportunity)
        - Only showcasing perfect outcomes (including challenges overcome builds more trust)
        
        ## Example Topics by Niche
        - Energy: "Zonnepanelen installateur: van 50 naar 200 leads per maand met lokale SEO"
        - Telecom: "Telecom provider: 89% minder churn door verbeterde klantenservice pagina's"
        - SaaS: "B2B SaaS: how we increased trial signups by 230% with content marketing"
        - E-commerce: "Webshop: 67% hogere conversie door productpagina optimalisatie"
        - Local services: "Aannemersbedrijf Rotterdam: van pagina 5 naar top 3 in Google Maps"
        
      • category-page.md 4 KB
        # Technique 32: E-Commerce Category Pages
        
        ## What It Is
        Navigational/commercial content that organizes products within a category, combining product listings with SEO-optimized category descriptions. The primary landing page for category-level searches. Captures "[category]", "[category] kopen", and "[category] vergelijken" searches.
        
        ## When to Use
        Commercial/navigational intent -- user wants to browse options within a product category. Every product category and meaningful subcategory needs its own page. These pages often rank for high-volume head terms.
        
        ## Structure Template
        
        ```
        H1: "[Category Name]"
            (e.g., "Zonnepanelen" or "Draadloze Koptelefoons")
        
        Category intro (above product grid):
            2-3 sentences: what the category includes, key buying criteria
            Filter/sort options prominent
        
        Product grid / listing:
            Product cards: image, name, price, rating, key spec
            Filterable by: price, brand, rating, key attributes
            Sortable by: price, popularity, rating, newest
        
        H2: [Category] koopgids (below product grid)
            3-5 paragraphs of unique, helpful category content
            Key criteria for choosing within this category
            Links to buying guide, comparisons, reviews
        
        H2: Popular subcategories
            Links to subcategory pages with brief descriptions
        
        H2: Veelgestelde vragen over [category]
            3-5 FAQ items specific to the category
        
        H2: Related categories
            Cross-links to adjacent categories
        ```
        
        ## Word Count
        300-1,000 words of unique content (excluding product listings -- the content supplements, not replaces, the product grid)
        
        ## Schema Markup
        - **Primary:** CollectionPage (with name, description)
        - **Secondary:** ItemList (for the product listing with ListItem entries)
        - **Tertiary:** BreadcrumbList (for navigation hierarchy)
        
        ## Featured Snippet Strategy
        - **Format:** List snippet for "types of [category]" queries
        - **Target:** The subcategories section or a bulleted list of product types
        - **Alternative:** Paragraph snippet for "what are [category]" queries via the intro text
        
        ## CTA Placement
        - **Primary:** Product cards with "bekijk" or "add to cart" per product
        - Filters and sorting: prominent, easy to use (improves engagement signals)
        - After category description: link to buying guide or comparison tool
        - Above fold: featured/promoted products or current deals
        
        ## Internal Linking Strategy
        - **Upward:** Link to parent category via breadcrumbs
        - **Downward:** Link to subcategory pages and individual product pages
        - **Sideways:** Link to related categories
        - **Supporting:** Link to buying guide, comparison articles, and how-to content
        - Receive links from pillar pages, navigation, and blog content
        
        ## Key Success Factors
        1. **Unique category content:** Do not rely solely on product listings -- add editorial content
        2. **Faceted navigation done right:** Filters must be crawlable but not create duplicate URL bloat
        3. **Above-the-fold products:** Product listings must be visible without scrolling past a wall of text
        4. **Breadcrumb navigation:** Clear hierarchy for both users and search engines
        5. **Pagination or load-more:** Handle large catalogs without harming crawlability
        6. **Category-specific filters:** A koptelefoon page filters on noise cancelling; a zonnepaneel page filters on wattage
        
        ## Common Mistakes
        - No unique content (just a product grid with no editorial value)
        - Too much content above the product grid (users came to browse products)
        - Duplicate content across similar category pages
        - Faceted navigation creating thousands of indexable thin pages
        - Missing breadcrumbs (confuses both users and crawlers)
        - No internal links to supporting content (buying guides, how-tos)
        - Broken pagination or infinite scroll that search engines cannot crawl
        
        ## Example Topics by Niche
        - Energy: "Zonnepanelen -- vergelijk en koop bij de beste aanbieders"
        - Telecom: "Sim only abonnementen -- vergelijk alle providers"
        - SaaS: "SEO Tools -- vergelijk de beste SEO software"
        - E-commerce: "Draadloze koptelefoons -- bekijk ons assortiment"
        - Local services: "Schilders in Amsterdam -- bekijk en vergelijk vakmensen"
        
      • comparison.md 5.4 KB
        # Technique 26: X vs Y Comparison Articles
        
        ## What It Is
        Commercial investigation content comparing two specific options side-by-side. Captures "[Product A] vs [Product B]" and "[A] of [B]" searches. One of the highest-converting content types in SEO.
        
        ## When to Use
        Commercial investigation intent -- user is deciding between two specific options they already know about. These users are close to a purchase decision.
        
        ## Structure Template (10-Section Framework)
        
        ```
        H1: "[Product A] vs [Product B]: [Outcome promise]"
            (e.g., "Vattenfall vs Eneco: welke is goedkoper in 2026?")
        
        H2: Quick verdict (above the fold)
            "Choose A if..." / "Choose B if..."
            Maximum 4 bullet points per option
        
        H2: At-a-glance comparison table
            4-6 key criteria in a clean table format
            Bold the winner per row
        
        H2: How we compared (methodology transparency)
            1-2 paragraphs explaining evaluation criteria and process
            E-E-A-T signal: shows expertise and transparency
        
        H2: Head-to-head by criteria
            H3: Criterion 1: [Category name]
                Verdict first: "[A] wins on [criterion] because..."
                Detailed analysis with specific data
            H3: Criterion 2: [Category name]
                Same pattern: verdict first, then evidence
            H3: Criterion 3-6: ...
        
        H2: Use cases
            3-5 real-world scenarios with recommendation per scenario
            "If you [situation], choose [option] because [reason]"
        
        H2: Pros and cons
            Specific, honest pros and cons for each option
            Not generic -- tied to real features and real limitations
        
        H2: Pricing comparison
            12-month total costs with gotchas/hidden fees
            Table format: plan names, monthly cost, annual cost, what's included
        
        H2: FAQ
            Pre-purchase questions, 2-4 sentences each
            Common: "Is [A] worth the extra cost?" "Can I switch from [A] to [B]?"
        
        H2: Final recommendation
            Clear recommendation with nuance per use case
            CTA to relevant conversion page
        ```
        
        ## Word Count
        1,200-2,500 words
        
        ## Schema Markup
        - **Primary:** FAQ (for FAQ section)
        - **Secondary:** Product (for both items compared)
        
        ## Featured Snippet Strategy
        - **Format:** Table snippet (comparison queries strongly trigger tables)
        - **Target:** The at-a-glance comparison table
        - **Alternative:** Paragraph snippet for the quick verdict
        
        ## CTA Placement
        - After quick verdict (above fold -- highest conversion point)
        - After final recommendation (bottom -- for readers who read through)
        - Inline within pricing section
        
        ## Internal Linking Strategy
        - Link to individual review articles for each product
        - Link to "best of" roundup for the category
        - Link to "alternatives to" pages for each product
        - Link to buying guide for the category
        - Receive links from pillar page and individual reviews
        
        ## Key Success Factors
        1. **Lead with the verdict:** Never make readers scroll to find which is better
        2. **Verdict-first per criterion:** Each H3 section starts with who wins and why
        3. **Consistent criteria:** Apply the exact same evaluation framework to both options
        4. **Real pricing with gotchas:** Include hidden fees, price increases, contract terms
        5. **Use case matching:** Specific scenarios ("If you're a family of 4 in an apartment...")
        6. **Honest limitations:** Acknowledge what each option does poorly (builds trust)
        7. **Methodology section:** How and why you compared these criteria (E-E-A-T signal)
        
        ## Common Mistakes
        - Not leading with the verdict (the #1 mistake)
        - Inconsistent evaluation criteria between products
        - No pricing transparency or missing total cost of ownership
        - Obviously biased toward one product without declaring it
        - Generic feature lists without real-world context
        - Missing the methodology section (undermines trust)
        - No use-case scenarios (readers can't self-select)
        
        ## Anti-AI Focus
        Evidence of actual usage is the strongest signal. Screenshots of dashboards, specific feature interactions, support experience anecdotes. AI can't fabricate these convincingly.
        
        To make comparison articles unmistakably experience-based:
        
        - **Include screenshots of both products.** Show the actual interface, dashboard, or configuration screen you encountered. Annotate screenshots to highlight differences. AI cannot generate authentic product screenshots.
        - **Describe specific interactions with support.** "When we contacted Eneco's support about a billing discrepancy, the response came within 2 hours via their app chat" is verifiable and specific in a way AI cannot replicate.
        - **Reference your testing methodology with dates.** "We signed up for both services in January 2026 and tracked costs over 3 months" establishes a timeline of actual usage.
        - **Note UI quirks, bugs, or friction points.** "The Vattenfall app crashes when switching between monthly and annual views on iOS 18" -- these micro-observations come only from real usage.
        - **Compare what the marketing says vs. what you experienced.** "Eneco advertises 'fixed pricing,' but the contract allows for a 5% annual adjustment" -- this kind of fine-print analysis signals thorough, hands-on evaluation.
        - **Include customer service response times and quality.** Document actual wait times, resolution quality, and communication style from real interactions.
        
        ## Example Topics by Niche
        - Energy: "Vattenfall vs Eneco vergelijken: prijs, service en duurzaamheid"
        - Telecom: "KPN vs Ziggo 2026: welke is beter voor glasvezel?"
        - SaaS: "HubSpot vs Salesforce: which CRM fits your business?"
        - E-commerce: "Shopify vs WooCommerce: wat is beter voor een webshop?"
        - Finance: "ABN AMRO vs ING hypotheek vergelijken"
        
      • definition.md 4.4 KB
        # Technique 19: "What Is" / Definition Articles
        
        ## What It Is
        Informational content that clearly defines a concept, term, or topic. Captures "what is [term]", "[term] meaning", and "[term] definition" searches.
        
        ## When to Use
        Informational intent -- user wants to understand a concept before taking action. High featured snippet opportunity. Strong top-of-funnel content for building topical authority.
        
        ## Structure Template
        
        ```
        H1: "What Is [Term]? [Clarifying subtitle]"
            (e.g., "Wat is een warmtepomp? Uitleg, kosten en voordelen")
        
        H2: Quick definition (above the fold)
            40-60 word paragraph optimized for featured snippet
            Plain language, no jargon in the first sentence
        
        H2: How [term] works
            Explain the mechanism or process
            Visual diagram or infographic recommended
        
        H2: Types of [term]
            H3: Type 1: [Name]
                Key characteristics, when it applies
            H3: Type 2: [Name]
            H3: Type 3: [Name]
        
        H2: Benefits / Why it matters
            Concrete advantages with data or examples
        
        H2: Drawbacks / Limitations
            Honest limitations (builds E-E-A-T trust)
        
        H2: [Term] in practice
            Real-world examples, case references, Dutch/European context
        
        H2: FAQ
            3-5 questions from People Also Ask data
        
        H2: Related topics
            Internal links to deeper articles
        ```
        
        ## Word Count
        1,000-2,000 words (enough to fully explain without padding)
        
        ## Schema Markup
        - **Primary:** DefinedTerm (with name and description)
        - **Secondary:** FAQ (for the FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** Paragraph snippet (definition queries strongly trigger paragraph snippets)
        - **Target:** The quick definition paragraph directly after H1
        - **Tip:** Start with "[Term] is..." to match Google's preferred snippet format
        
        ## CTA Placement
        - Mid-article (after "how it works"): related tool, calculator, or comparison
        - End of article: deeper guide, consultation, or relevant product page
        
        ## Internal Linking Strategy
        - Link to how-to articles for applying the concept
        - Link to comparison articles for choosing between types
        - Link to pillar page for the broader topic cluster
        - Link to glossary page where this term also appears
        - Receive links from how-tos, guides, and FAQ pages that reference the term
        
        ## Key Success Factors
        1. **First sentence is the definition:** No preamble -- start with "[Term] is..." immediately
        2. **Plain language first, technical second:** Define in simple words, then add nuance
        3. **Visual explanation:** Diagrams or infographics dramatically increase dwell time
        4. **Types/categories section:** Users often search "types of [term]" as a follow-up
        5. **Practical context:** Show how the concept applies in the Dutch/European market
        6. **Keep it focused:** One concept per article -- do not blend multiple definitions
        
        ## Common Mistakes
        - Long introductions before the actual definition (loses snippet and reader)
        - Circular definitions ("SEO is the practice of doing SEO")
        - Too academic or jargon-heavy for the target audience
        - Missing the "types" section (leaves related search intent unaddressed)
        - No practical examples or real-world context
        - Forgetting to link to action-oriented content (how-tos, buying guides)
        
        ## Anti-AI Focus
        Definition articles are frequently targeted by AI content generators because the format is straightforward. To differentiate:
        
        - **Ground definitions in local context.** Reference specific Dutch or European regulations, market conditions, or cultural norms that generic AI-generated definitions miss. For example, when defining "salderen," include the current Dutch net metering rules and their planned phase-out timeline.
        - **Use real-world examples from your domain.** Instead of textbook examples, reference actual companies, products, or situations your audience would recognize.
        - **Include nuance and edge cases.** AI definitions tend to be clean and tidy. Real expertise means knowing the exceptions, the "it depends" scenarios, and the contexts where the standard definition breaks down.
        - **Add the "in practice" section with genuine observation.** Describe how the concept actually plays out in the market, including surprises or counterintuitive behaviors you have observed firsthand.
        
        ## Example Topics by Niche
        - Energy: "Wat is salderen? Uitleg en regels voor zonnepanelen"
        - Telecom: "Wat is glasvezel? Verschil met kabel en DSL uitgelegd"
        - SaaS: "What is a CDP? Customer Data Platform explained"
        - E-commerce: "Wat is dropshipping? Zo werkt het in Nederland"
        - Local services: "Wat is een VvE? Uitleg voor appartementseigenaren"
        
      • faq-page.md 4.7 KB
        # Technique 21: FAQ Pages
        
        ## What It Is
        Structured question-and-answer content that addresses the most common questions around a topic, product, or service. Captures long-tail "question" queries and People Also Ask opportunities.
        
        ## When to Use
        Informational intent -- user has specific questions and wants direct answers. Use when PAA data shows a cluster of related questions, or when a topic has many common questions that do not justify individual articles.
        
        ## Structure Template
        
        ```
        H1: "[Topic] FAQ: Veelgestelde vragen over [topic]"
            (e.g., "Zonnepanelen FAQ: 25 veelgestelde vragen beantwoord")
        
        H2: Short intro
            1-2 sentences explaining what this page covers
            Link to pillar page for full overview
        
        H2: [Question category 1]
            H3: [Question 1]?
                2-4 sentence answer, front-load the key fact
            H3: [Question 2]?
                Same pattern
            H3: [Question 3-5]?
        
        H2: [Question category 2]
            H3: [Question 6]?
            H3: [Question 7-10]?
        
        H2: [Question category 3]
            H3: [Question 11-15]?
        
        H2: [Question category 4: Costs/Pricing questions]
            H3: [Cost question 1]?
            H3: [Cost question 2]?
        
        H2: Still have questions?
            CTA to contact, consultation, or community
        ```
        
        ## Word Count
        1,500-2,500 words (15-25 questions at 60-120 words per answer)
        
        ## Schema Markup
        - **Primary:** FAQPage (with Question and acceptedAnswer pairs)
        - **Secondary:** BreadcrumbList
        
        ## Featured Snippet Strategy
        - **Format:** Paragraph snippet (FAQ queries trigger paragraph snippets)
        - **Target:** Each H3 question with its concise answer paragraph
        - **Tip:** Start each answer with the direct answer in the first sentence, then elaborate
        
        ## CTA Placement
        - After each question category: contextual link to relevant deeper content
        - End of page: contact form, consultation CTA, or chat prompt
        - Inline within answers: link to products/services when naturally relevant
        
        ## Internal Linking Strategy
        - Link from answers to detailed articles that cover the topic in depth
        - Link to definition articles for technical terms mentioned in answers
        - Link to how-to articles for process-related questions
        - Link to comparison/review pages for product-related questions
        - Receive links from pillar pages and service pages
        
        ## Key Success Factors
        1. **Real questions from real data:** Use PAA, search console queries, and customer support data
        2. **Answer-first format:** Every answer starts with the direct answer, then adds context
        3. **Group by category:** Do not list 25 random questions -- organize by theme
        4. **Keep answers concise:** 60-120 words per answer. Link to deeper content for complex topics
        5. **Update regularly:** Add new questions based on fresh PAA and support ticket data
        6. **Use the exact question phrasing:** Match how users actually search, not how you would phrase it
        
        ## Common Mistakes
        - Using made-up questions instead of real search queries
        - Answers that are too long (200+ words) -- defeats the purpose of an FAQ
        - Answers that are too short (one sentence) -- provides no value and looks thin
        - No logical grouping of questions (random order confuses readers)
        - Missing FAQPage schema markup (loses AI Overview citation lift — ~3x per vendor studies; Google still uses it for entity understanding)
        - Duplicating content from other pages instead of summarizing and linking
        - Not updating after launch (FAQ pages go stale quickly)
        
        ## Anti-AI Focus
        FAQ pages are a common target for AI generation because the Q&A format is easy to produce at scale. To ensure your FAQ pages signal genuine expertise:
        
        - **Source questions from real user data.** Pull questions directly from Google Search Console queries, customer support tickets, live chat logs, and sales call transcripts. AI-generated FAQs tend to use obvious, generic questions.
        - **Include answers that reflect actual customer interactions.** Reference specific situations, edge cases, or follow-up questions that real customers have raised. "We get asked this a lot after customers try X and find that Y happens" is a pattern AI cannot fabricate.
        - **Add specificity to answers.** Instead of "prices vary," state "as of March 2026, prices range from X to Y depending on Z." Concrete numbers, dates, and conditions are strong authenticity signals.
        - **Update with real frequency data.** Note which questions are most common ("this is our #1 support question") or trending ("we started getting this question after the 2026 regulation change").
        
        ## Example Topics by Niche
        - Energy: "Veelgestelde vragen over overstappen van energieleverancier"
        - Telecom: "Glasvezel FAQ: alles wat je wilt weten over glasvezel internet"
        - SaaS: "Frequently asked questions about CRM implementation"
        - E-commerce: "FAQ retourneren en ruilen: alles over ons retourbeleid"
        - Local services: "Verhuizen FAQ: veelgestelde vragen over verhuisservice"
        
      • glossary-page.md 4.9 KB
        # Technique 25: Glossary / Definition Hub Pages
        
        ## What It Is
        A centralized collection of short definitions for industry-specific terms, organized alphabetically or by category. Each term gets a concise explanation. Functions as a topical authority signal and internal linking hub.
        
        ## When to Use
        Informational intent -- user wants quick definitions of industry jargon. Use when your niche has 30+ specialized terms that users search for. The glossary page acts as a hub; high-value terms also get dedicated definition articles.
        
        ## Structure Template
        
        ```
        H1: "[Industry] Woordenlijst: [X] Begrippen Uitgelegd"
            (e.g., "Energie woordenlijst: 80 begrippen over stroom, gas en verduurzaming")
        
        H2: Introduction
            1-2 sentences: what this glossary covers and who it's for
            Anchor link navigation: A-Z or by category
        
        H2: A
            H3: [Term 1]
                2-4 sentence definition
                Link to full article if one exists
            H3: [Term 2]
        
        H2: B
            H3: [Term 3]
            H3: [Term 4]
        
        ... (continue alphabetically or by category)
        
        H2: [Category-based alternative]
            H3: Basic concepts
            H3: Technical terms
            H3: Financial terms
            H3: Legal/regulatory terms
        ```
        
        ## Word Count
        200-400 words per term (total page length depends on number of terms -- typically 5,000-15,000 words for a full glossary)
        
        ## Schema Markup
        - **Primary:** DefinedTermSet (with individual DefinedTerm entries)
        - **Secondary:** BreadcrumbList
        
        ## Featured Snippet Strategy
        - **Format:** Paragraph snippet (definition queries trigger paragraph snippets)
        - **Target:** Each H3 term heading with its concise definition paragraph
        - **Tip:** Start each definition with "[Term] is..." for maximum snippet eligibility
        
        ## CTA Placement
        - Sticky navigation or sidebar: search/filter functionality for finding terms
        - Within definitions: contextual links to related products, tools, or services
        - End of page: pillar page CTA or newsletter signup for industry updates
        
        ## Internal Linking Strategy
        - **Outbound hub:** Every term links to its dedicated definition article (if one exists)
        - **Outbound to related:** Terms cross-reference other glossary entries
        - **Receive links:** Every article in the cluster links technical terms back to the glossary
        - **Anchor links:** Use #term-name anchors so other pages can deep-link to specific definitions
        
        ## Key Success Factors
        1. **Alphabetical AND searchable:** Provide both A-Z navigation and on-page search/filter
        2. **Concise definitions:** 200-400 words maximum per term -- this is a reference, not an article
        3. **Link to deep-dives:** High-value terms should have their own dedicated definition article
        4. **Cross-reference terms:** Link related terms within definitions (builds internal link web)
        5. **Keep it current:** Add new terms as the industry evolves, remove obsolete ones
        6. **Anchor links per term:** Enable deep-linking from any page to a specific glossary entry
        
        ## Common Mistakes
        - Definitions that are too long (turns the glossary into an encyclopedia)
        - Definitions that are too short (one-sentence definitions provide no value)
        - No navigation system (100+ terms without A-Z anchors is unusable)
        - Not linking to dedicated articles for important terms
        - Using jargon to define jargon (circular definitions)
        - Publishing once and never updating
        - Missing DefinedTerm schema (loses rich result opportunity for definitions)
        
        ## Anti-AI Focus
        Glossary pages are frequently AI-generated because the format is repetitive and the content seems straightforward. To differentiate your glossary from AI-produced alternatives:
        
        - **Write definitions from the practitioner's perspective.** Instead of textbook definitions, explain terms the way you would to a new colleague. Include the practical implications that matter in your specific industry context.
        - **Add "why this matters" context.** After the definition, briefly note why this term is relevant to your audience. "This is important because in the Dutch market, salderen rules are changing in 2027" adds value AI glossaries lack.
        - **Include common misunderstandings.** For terms that are frequently confused or misused, note the distinction. "Often confused with X, but the key difference is..." demonstrates expertise.
        - **Cross-reference with specificity.** When linking related terms, explain the relationship: "See also: teruglevering (which covers the actual process of feeding electricity back to the grid, whereas salderen refers to the financial offsetting)."
        - **Update with industry evolution.** Add notes when term definitions shift due to new regulations, technology changes, or market developments, with dates.
        
        ## Example Topics by Niche
        - Energy: "Energie woordenlijst: van aardwarmte tot zonnepaneel -- 80 begrippen uitgelegd"
        - Telecom: "Telecom woordenboek: 60 internetbegrippen eenvoudig uitgelegd"
        - SaaS: "Marketing glossary: 100 digital marketing terms defined"
        - E-commerce: "E-commerce woordenlijst: 50 begrippen die elke webshop-eigenaar moet kennen"
        - Local services: "Bouwterminologie: 70 bouwkundige begrippen voor opdrachtgevers"
        
      • how-to.md 4.6 KB
        # Technique 18: How-To / Tutorial Articles
        
        ## What It Is
        Informational content that guides users through accomplishing a specific task. Captures "how to [action]" and "[action] tutorial" searches.
        
        ## When to Use
        Informational intent -- user wants to accomplish a specific task. High PAA presence, strong featured snippet opportunity.
        
        ## Structure Template
        
        ```
        H1: "How to [Action] [Qualifier]"
            (e.g., "Hoe je zonnepanelen installeert in 2026")
        
        H2: Quick answer / TL;DR
            40-60 words, optimized for featured snippet (paragraph or list format)
        
        H2: What you need / Prerequisites
            Materials, tools, knowledge required
        
        H2: Step-by-step instructions
            H3: Step 1: [Action verb] [specific outcome]
                Visual per step (screenshot, photo, diagram)
            H3: Step 2: ...
            H3: Step N: ...
        
        H2: Common mistakes to avoid
            Practical, specific mistakes (not generic warnings)
        
        H2: Tips from experience
            First-person insights that demonstrate E-E-A-T
        
        H2: FAQ
            3-5 questions from People Also Ask data
        
        H2: Next steps / Related guides
            Internal links to related content
        ```
        
        ## Word Count
        1,500-3,000 words (intent-dependent -- simple tasks shorter, complex tasks longer)
        
        ## Schema Markup
        - **Primary:** HowTo (with step-by-step structured data)
        - **Secondary:** FAQ (for the FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** Ordered list snippet (46.91% of "how" queries)
        - **Target:** Place numbered steps as an ordered list after the relevant H2
        - **Quick answer:** 40-60 word paragraph answer directly after H1 for paragraph snippet
        
        ## CTA Placement
        - Mid-article (after step 3-4): related tool, product, or service
        - End of article: deeper guide, consultation, or conversion page
        
        ## Internal Linking Strategy
        - Link to definition pages for technical jargon
        - Link to related how-tos (next steps)
        - Link to relevant product/service pages for tools mentioned
        - Link to pillar page for the topic cluster
        - Receive links from pillar page and FAQ hubs
        
        ## Key Success Factors
        1. **Front-load the answer:** The quick answer section must appear before the detailed steps
        2. **One action per step:** Each step should describe exactly one thing the reader does
        3. **Visual evidence:** Screenshots, photos, or diagrams per step demonstrate Experience
        4. **Specific measurements:** "Drill a 6mm hole 3cm from the edge" not "drill a hole"
        5. **Common mistakes are gold:** These differentiate from generic guides and signal experience
        
        ## Common Mistakes
        - Burying the answer below a long introduction
        - Missing the quick-answer paragraph (loses featured snippet opportunity)
        - Not including visuals per step (text-only how-tos underperform)
        - Generic steps that could apply to any topic (Horoscope Test failure)
        - Missing prerequisites section (frustrates readers mid-process)
        
        ## Anti-AI Focus
        Most effective anti-AI signal for tutorials: mention specific error messages, version numbers, and unexpected behaviors that only someone who actually did this would know.
        
        How-to articles are one of the content types most vulnerable to AI-generated imitation because the step-by-step format is easy to replicate structurally. What AI cannot replicate is the texture of real experience. To make your tutorials unmistakably human-authored:
        
        - **Include specific error messages and edge cases.** When you describe a step, mention the exact error a user might encounter ("If you see 'Error 0x80070005: Access Denied', right-click and run as administrator"). Only someone who has actually performed the process knows these details.
        - **Reference version numbers and dates.** State the exact software version, firmware revision, or product model you used. AI tends to generalize; specificity signals authenticity.
        - **Document unexpected behaviors.** Describe what happened that the official documentation does not mention -- the step that took longer than expected, the setting that was in a different location than described, the workaround you had to improvise.
        - **Show your environment.** Include screenshots of your actual screen, photos of your workspace, or terminal output with timestamps. These artifacts are nearly impossible to fabricate convincingly.
        - **Mention what did NOT work.** Describing failed approaches before arriving at the working solution is a strong experience signal. AI-generated content almost never includes false starts or dead ends.
        
        ## Example Topics by Niche
        - Energy comparison: "Hoe kies je de beste energieleverancier"
        - Local services: "Hoe vind je een betrouwbare loodgieter in Amsterdam"
        - SaaS: "How to set up automated email workflows"
        - E-commerce: "Hoe retourneer je een bestelling bij Bol.com"
        - Finance: "Hoe vraag je hypotheekrenteaftrek aan"
        
      • integration-page.md 3.9 KB
        # Technique 35: SaaS Integration Pages
        
        ## What It Is
        Product pages dedicated to a specific integration between your software and a third-party tool. Captures "[your product] + [partner tool] integration", "[tool A] koppelen aan [tool B]", and "[tool] API" searches. Scalable SEO strategy -- one page per integration.
        
        ## When to Use
        Commercial/navigational intent -- user wants to know if your product works with a tool they already use, or is looking for a solution that connects two specific tools. High conversion potential because it addresses a specific compatibility need.
        
        ## Structure Template
        
        ```
        H1: "[Your Product] + [Partner Tool] Integratie"
            (e.g., "InhouseSEO + Google Search Console integratie")
        
        Hero section:
            Both product logos side by side
            1-sentence value proposition of the integration
            CTA: "Connect now" or "Start free trial"
        
        H2: What this integration does
            3-5 key capabilities as bullet points
            Focus on outcomes, not technical details
        
        H2: Key features
            H3: [Feature 1]: [Benefit]
                Screenshot showing the integration in action
            H3: [Feature 2]: [Benefit]
            H3: [Feature 3]: [Benefit]
        
        H2: How to set it up
            3-5 steps with screenshots
            Estimated setup time
            Prerequisites (accounts needed, permissions, etc.)
        
        H2: Use cases
            H3: [Use case 1]
                Scenario + how the integration helps
            H3: [Use case 2]
            H3: [Use case 3]
        
        H2: FAQ
            "Is the integration free?"
            "What data is synced?"
            "How often does it sync?"
            "Is my data secure?"
        
        H2: Related integrations
            Links to other integration pages
        ```
        
        ## Word Count
        800-1,500 words (focused and practical -- users want compatibility confirmation and setup instructions)
        
        ## Schema Markup
        - **Primary:** SoftwareApplication (name, applicationCategory, operatingSystem)
        - **Secondary:** HowTo (for the setup steps)
        
        ## Featured Snippet Strategy
        - **Format:** List snippet for "how to connect [A] to [B]" queries
        - **Target:** The setup steps section
        - **Alternative:** Paragraph snippet for "does [A] integrate with [B]" queries
        
        ## CTA Placement
        - Hero section: primary CTA above the fold
        - After features: "try this integration" CTA
        - After setup steps: "connect now" CTA
        - End of page: trial or demo CTA
        
        ## Internal Linking Strategy
        - Link to integrations hub/directory page
        - Link to related integration pages (same category of tools)
        - Link to how-to articles for advanced configuration
        - Link to pricing page (which plans include this integration)
        - Receive links from integrations directory, partner's integration directory, and blog content
        
        ## Key Success Factors
        1. **Both logos prominently displayed:** Visual confirmation of the partnership/compatibility
        2. **Setup instructions included:** Users need to know how easy (or hard) it is before committing
        3. **Screenshots of the integration working:** Visual proof that it actually works
        4. **Specific data flow:** What data moves between systems, how often, in which direction
        5. **One page per integration:** Do not combine multiple integrations on one page
        6. **Partner tool SEO:** Include the partner tool's name in title, URL, and headings
        
        ## Common Mistakes
        - Vague description of what the integration does (be specific about data and actions)
        - No setup instructions (forces users to sign up before knowing the process)
        - No screenshots (users want visual proof before committing)
        - Combining multiple integrations on one page (dilutes SEO value)
        - Not mentioning which pricing plans include the integration
        - Missing the FAQ section (integration pages generate specific technical questions)
        - Not linking from the partner's directory or marketplace
        
        ## Example Topics by Niche
        - Energy: "Toon slimme thermostaat + Google Home koppeling"
        - Telecom: "Ziggo Mediabox + Chromecast integratie instellen"
        - SaaS: "InhouseSEO + Google Search Console: automatische data-import"
        - E-commerce: "Shopify + Mollie betaalintegratie: zo stel je het in"
        - Local services: "Werkspot + Google Agenda koppeling voor vakmensen"
        
      • landing-page.md 4.4 KB
        # Technique 33: Landing Pages
        
        ## What It Is
        Conversion-focused pages designed to turn visitors into leads or customers. Minimal navigation, single clear CTA, and persuasive copy. Can target both organic search and paid traffic. Captures "[service] aanvragen", "[tool] proberen", and "[action] offerte" searches.
        
        ## When to Use
        Transactional intent -- user is ready to take action (sign up, request a quote, start a trial, download). Use for campaigns, specific offers, or high-intent keywords where the goal is conversion, not education.
        
        ## Structure Template
        
        ```
        H1: "[Clear value proposition in one line]"
            (e.g., "Bespaar tot 30% op je energierekening -- vergelijk nu gratis")
        
        Hero section:
            H1 + supporting subtitle (1 sentence)
            Primary CTA button (contrasting color, action-oriented text)
            Trust signals: customer logos, rating, "X klanten gingen je voor"
            Hero image or product screenshot
        
        H2: [Problem statement]
            2-3 sentences describing the pain point
            "Herkenbaar?" or "Klinkt dit bekend?"
        
        H2: How [product/service] solves this
            3-4 key benefits as icon + headline + 1-sentence description
            Benefits, not features -- focus on outcomes
        
        H2: How it works
            3 steps: simple process visualization
            Step 1: [Action] → Step 2: [Action] → Step 3: [Result]
        
        H2: Social proof
            Testimonials (2-3, with name, photo, company)
            Key metric: "4.8/5 based on 2,340 reviews"
            Customer logos or media mentions
        
        H2: [Objection handling]
            Address top 3 objections directly
            "No credit card required" / "Cancel anytime" / "GDPR compliant"
        
        CTA section (repeated):
            Same primary CTA as hero
            Urgency or scarcity element if genuine
        
        H2: FAQ (optional, below fold)
            3-5 conversion-related questions
        ```
        
        ## Word Count
        800-1,200 words (every word must serve the conversion goal -- ruthlessly cut anything that does not)
        
        ## Schema Markup
        - **Primary:** WebPage or Product/Service (depending on what is offered)
        - **Secondary:** FAQ (if FAQ section is included)
        - **Tertiary:** Organization (trust signals)
        
        ## Featured Snippet Strategy
        - **Format:** Not the primary goal for landing pages
        - **Target:** Focus on ranking for transactional keywords, not snippet capture
        - **Tip:** The H1 and meta description should include the primary transactional keyword
        
        ## CTA Placement
        - **Hero section:** Primary CTA above the fold (most important placement)
        - **After benefits:** Reinforcement CTA
        - **After social proof:** Strongest conversion point -- repeat primary CTA
        - **Sticky bar:** Persistent CTA on scroll (mobile and desktop)
        - **Maximum 1 CTA type:** Do not offer multiple different actions -- one page, one goal
        
        ## Internal Linking Strategy
        - **Minimal outbound links:** Do not link away from the page unnecessarily
        - Link to terms, privacy, and legal pages in footer only
        - Receive links from blog content, email campaigns, and paid traffic
        - Receive links from related service or product pages
        - Do not include main site navigation if possible (focused experience)
        
        ## Key Success Factors
        1. **One page, one goal:** Every element must serve the single conversion objective
        2. **Above-the-fold CTA:** The primary CTA must be visible without scrolling
        3. **Social proof is non-negotiable:** Testimonials, ratings, or customer counts build trust
        4. **Address objections explicitly:** "No credit card", "Cancel anytime", "Free trial"
        5. **Fast load time:** Landing pages must load in under 2 seconds (conversion drops 7% per second)
        6. **Mobile-first design:** Majority of traffic is mobile -- design for thumb-friendly CTAs
        
        ## Common Mistakes
        - Too many CTAs or competing actions (confuses the visitor)
        - Wall of text without visual hierarchy (no one reads landing pages linearly)
        - No social proof (claims without evidence do not convert)
        - Including full site navigation (provides escape routes from conversion)
        - Generic stock photos instead of real product screenshots or customer photos
        - CTA text that says "Submit" instead of action-oriented text ("Start gratis", "Vergelijk nu")
        - Not testing variations (landing pages should always be A/B tested)
        
        ## Example Topics by Niche
        - Energy: "Energievergelijker -- bespaar tot 30% op stroom en gas"
        - Telecom: "Glasvezel aanvragen -- check beschikbaarheid op jouw adres"
        - SaaS: "Start your free 14-day trial -- no credit card required"
        - E-commerce: "Gratis verzending op je eerste bestelling -- meld je nu aan"
        - Local services: "Ontvang 3 gratis offertes van schilders in jouw regio"
        
      • listicle.md 5.4 KB
        # Technique 27: "Best X for Y" Listicle Roundups
        
        ## What It Is
        Curated list content ranking or recommending multiple options within a category for a specific audience or use case. Captures "best [product/service] for [audience/use case]" searches. High commercial intent with strong conversion potential.
        
        ## When to Use
        Commercial investigation intent -- user is exploring options within a category but has not narrowed down to specific products. These searches signal active buying consideration and respond well to structured comparisons.
        
        ## Structure Template
        
        ```
        H1: "Best [Category] for [Audience/Use Case] in [Year]"
            (e.g., "Beste energieleveranciers van 2026: top 10 vergeleken")
        
        H2: Our top picks at a glance
            Summary table with top 3-5 picks and key differentiator
            "Best overall", "Best budget", "Best for [specific need]"
        
        H2: How we evaluated
            Criteria used, methodology, any testing performed
            E-E-A-T signal: demonstrates expertise and transparency
        
        H2: 1. [Top pick name] -- Best overall
            H3: Overview
            H3: Key features / What we like
            H3: Drawbacks
            H3: Pricing
            H3: Best for: [specific use case]
        
        H2: 2. [Second pick] -- Best for [specific angle]
            Same structure as above
        
        H2: 3-10. [Remaining picks]
            Same structure, can be slightly shorter for lower-ranked options
        
        H2: Honorable mentions
            2-3 options that narrowly missed the list
        
        H2: How to choose the right [category]
            Decision framework: 3-5 questions to ask yourself
            Links to buying guide
        
        H2: FAQ
            5-7 pre-purchase questions
        
        H2: Methodology
            Detailed explanation of how options were tested/evaluated
        ```
        
        ## Word Count
        2,000-4,000 words (depends on number of items -- 200-400 words per item plus intro/outro)
        
        ## Schema Markup
        - **Primary:** ItemList (with ListItem entries and position)
        - **Secondary:** FAQ (for FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** List snippet (listicle queries strongly trigger numbered lists)
        - **Target:** The "top picks at a glance" section
        - **Alternative:** Table snippet for the summary comparison table
        
        ## CTA Placement
        - After each item: affiliate link, "check price", or "visit site" button
        - After top 3: comparison tool or quiz ("not sure which? take our quiz")
        - End of article: buying guide link or consultation CTA
        
        ## Internal Linking Strategy
        - Link to individual review articles for each listed item
        - Link to "X vs Y" comparison articles for top picks
        - Link to buying guide for the category
        - Link to "alternatives to" pages for individual items
        - Receive links from pillar page, individual reviews, and definition articles
        
        ## Key Success Factors
        1. **Lead with top picks:** Summary table above the fold -- do not make readers scroll through 10 items
        2. **Consistent evaluation framework:** Same criteria applied to every item in the same order
        3. **Specific "best for" labels:** Each item should win in a specific category or use case
        4. **Real drawbacks:** Including honest negatives builds trust more than all-positive reviews
        5. **Methodology section:** Explain how you evaluated -- testing, criteria weights, data sources
        6. **Update quarterly:** Pricing, features, and rankings change -- stale listicles lose trust and rankings
        7. **Limit the list:** 7-10 items is optimal. 20+ items overwhelms and suggests no actual curation
        
        ## Common Mistakes
        - No summary table at the top (readers want the answer fast)
        - All items described equally positively (no differentiation = no value)
        - Missing methodology section (undermines credibility)
        - Too many items without clear ranking criteria
        - Not disclosing affiliate relationships where applicable
        - Outdated pricing or features (loses trust immediately)
        - Generic descriptions that could apply to any competitor
        
        ## Anti-AI Focus
        Listicle roundups are one of the most commonly AI-generated content types because the format is formulaic. To ensure yours stands apart:
        
        - **Base rankings on documented evaluation.** Describe the specific testing process: "We signed up for all 10 providers, used each for at least 2 weeks, and scored them on 6 criteria." AI cannot claim a testing process it never performed.
        - **Include specific, dated observations.** "As of February 2026, Provider X's onboarding flow takes 4 steps and about 3 minutes" -- this level of specificity signals hands-on experience.
        - **Show why items are ranked in this order.** Do not just list features; explain the trade-offs that led to your ranking. "We ranked Y above Z despite Z's lower price because Y's customer support resolved issues in under an hour during our testing."
        - **Include items you intentionally excluded and why.** "We considered Provider W but excluded it because their pricing changed mid-review and the new terms were not competitive." This editorial judgment is distinctly human.
        - **Add personal recommendations per reader profile.** "If you are a freelancer working from home, our #4 pick is actually better for you than our #1" -- this kind of specific advice requires real understanding of the products and the audience.
        
        ## Example Topics by Niche
        - Energy: "Beste energieleveranciers 2026: top 10 op prijs, service en duurzaamheid"
        - Telecom: "Beste internet providers Nederland 2026: vergeleken op snelheid en prijs"
        - SaaS: "Best project management tools for small teams in 2026"
        - E-commerce: "Beste webshop platforms 2026: Shopify, WooCommerce en 8 alternatieven"
        - Local services: "Beste hypotheekadviseurs 2026: top 10 vergeleken op kosten en reviews"
        
      • location-page.md 4.5 KB
        # Technique 37: Location / Service Area Pages
        
        ## What It Is
        Locally-optimized content targeting geographic-specific searches for services or businesses. Each page focuses on one specific location or service area. Captures "[service] [city]", "[service] in de buurt", and "[service] [region]" searches. Critical for local SEO and Google Maps visibility.
        
        ## When to Use
        Local/transactional intent -- user wants a service or business in a specific geographic area. Create one page per location you serve. Must be genuinely unique per location -- Google penalizes doorway pages with only the city name swapped.
        
        ## Structure Template
        
        ```
        H1: "[Service] in [Location]"
            (e.g., "Zonnepanelen installatie in Rotterdam")
        
        Hero section:
            H1 + location-specific value proposition
            CTA: "Vraag een gratis offerte aan in [Location]"
            Trust: "X klanten in [Location]" or local rating
        
        H2: [Service] in [Location]: wat u moet weten
            Location-specific context (regulations, climate, building types)
            Why [Location] is unique for this service
        
        H2: Onze diensten in [Location]
            Services offered in this specific area
            Any location-specific variations or specialties
        
        H2: Waarom kiezen voor ons in [Location]?
            Local expertise, nearby team, response times
            Location-specific credentials or partnerships
        
        H2: Recente projecten in [Location]
            2-3 local case studies or project examples
            Photos of actual work in the area
            Specific addresses or neighborhoods (with permission)
        
        H2: Prijsindicatie voor [Location]
            Location-specific pricing factors
            Local market rates or ranges
            Why prices may differ from other areas
        
        H2: Werkgebied en bereikbaarheid
            Map embed showing service area
            Neighborhoods, districts, or postal codes served
            Response time for this area
        
        H2: Veelgestelde vragen over [service] in [Location]
            Location-specific FAQ (local regulations, permits, etc.)
            3-5 questions
        
        H2: Contact in [Location]
            Local address (if applicable)
            Phone number with local area code
            Contact form
            Google Maps embed
        ```
        
        ## Word Count
        1,200-1,500 words (substantial enough to be genuinely useful, not so long that it's clearly padded)
        
        ## Schema Markup
        - **Primary:** LocalBusiness (name, address, telephone, geo, openingHours)
        - **Secondary:** Service (serviceType, areaServed, provider)
        - **Tertiary:** FAQ (for the FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** Paragraph snippet for "[service] [location]" queries
        - **Target:** The opening section describing the service in context of the location
        - **Tip:** Include the location name naturally in the first 50 words
        
        ## CTA Placement
        - **Hero section:** Location-specific CTA above the fold
        - **After local projects:** Social proof reinforcement CTA
        - **After pricing:** "Get a quote for [Location]" CTA
        - **End of page:** Full contact form with address/phone
        
        ## Internal Linking Strategy
        - Link to main service page for detailed service information
        - Link to nearby location pages ("Ook actief in [nearby city]")
        - Link to relevant case studies from the area
        - Link to pricing page for detailed pricing
        - Receive links from main service page, city hub page, and blog content
        
        ## Key Success Factors
        1. **Genuinely unique content per location:** Each page must have location-specific information, not just city-name swaps
        2. **Local case studies or projects:** Real examples from the area prove local presence
        3. **Location-specific data:** Local pricing, regulations, permits, climate factors
        4. **NAP consistency:** Name, Address, Phone must match Google Business Profile exactly
        5. **Google Maps embed:** Visual confirmation of location and service area
        6. **Local schema markup:** Complete LocalBusiness schema with geo coordinates
        
        ## Common Mistakes
        - **Doorway pages:** Identical content with only the city name changed (Google penalty risk)
        - No local case studies or project examples (no proof of local presence)
        - Missing LocalBusiness schema (loses local pack eligibility)
        - NAP inconsistency between the page and Google Business Profile
        - No embedded map (missed engagement signal and trust element)
        - Creating pages for locations you do not actually serve
        - Not including location-specific pricing or regulatory information
        
        ## Example Topics by Niche
        - Energy: "Zonnepanelen installatie Rotterdam -- lokale installateurs en prijzen"
        - Telecom: "Glasvezel beschikbaarheid Amsterdam -- check per postcode"
        - SaaS: "SEO bureau Utrecht -- lokaal online marketing advies"
        - E-commerce: "Same-day delivery Amsterdam -- bestel voor 14:00"
        - Local services: "Loodgieter Den Haag -- 24/7 spoed en onderhoud"
        
      • news-article.md 5 KB
        # Technique 23: News / Trend Analysis Articles
        
        ## What It Is
        Time-sensitive content covering breaking news, industry developments, or emerging trends. Captures trending queries, Google News inclusion, and Discover traffic. Speed of publication is the primary ranking factor.
        
        ## When to Use
        Informational/navigational intent -- user wants to understand a recent event or emerging trend. Use when there is a newsworthy development, regulatory change, or market shift. First-mover advantage is significant.
        
        ## Structure Template
        
        ```
        H1: "[News event]: [Impact or what it means]"
            (e.g., "Nieuwe energiewet 2026: wat verandert er voor consumenten?")
        
        H2: Key takeaways
            3-5 bullet points summarizing the essential facts
            Optimized for featured snippet and Discover
        
        H2: What happened
            Factual summary of the event or announcement
            Who, what, when, where, why
            Source attribution
        
        H2: Why it matters
            Impact analysis for the target audience
            Specific consequences with data or examples
        
        H2: What to expect next
            Short-term and medium-term implications
            Expert quotes or industry reactions if available
        
        H2: What you should do now
            Actionable steps for the reader
            Links to relevant tools, guides, or services
        
        H2: Background / Context
            Brief history for readers unfamiliar with the topic
            Link to pillar page or deeper explainer
        ```
        
        ## Word Count
        800-1,500 words (speed trumps comprehensiveness -- publish fast, expand later)
        
        ## Schema Markup
        - **Primary:** NewsArticle (with datePublished, dateModified, author)
        - **Secondary:** FAQ (if adding FAQ section in updates)
        
        ## Featured Snippet Strategy
        - **Format:** Paragraph snippet for "what happened" queries
        - **Target:** The opening "key takeaways" section or the first paragraph of "what happened"
        - **Tip:** Use the inverted pyramid -- most important information first
        
        ## CTA Placement
        - After "what you should do": relevant tool, service, or comparison page
        - End of article: newsletter signup for future updates on this topic
        - Inline: links to related evergreen content
        
        ## Internal Linking Strategy
        - Link to evergreen pillar page for topic background
        - Link to relevant how-to articles for action steps
        - Link to previous news articles in the same topic (creates a news trail)
        - Update pillar pages and statistics pages to reference this news
        - Receive links from social shares, newsletters, and news aggregators
        
        ## Key Success Factors
        1. **Publish speed:** First to publish with accurate information wins -- aim for hours, not days
        2. **Inverted pyramid:** Most critical information in the first 100 words
        3. **Fact-check rigorously:** Speed cannot sacrifice accuracy -- verify before publishing
        4. **Update the article:** Add developments, reactions, and corrections with visible timestamps
        5. **Original analysis:** Do not just rewrite the press release -- add impact analysis and context
        6. **Expand later:** Publish a short version fast, then expand with analysis within 24-48 hours
        
        ## Common Mistakes
        - Prioritizing word count over speed (800 accurate words now beats 2,000 words tomorrow)
        - Publishing without fact-checking (corrections damage E-E-A-T)
        - No "what it means for you" section (readers need impact, not just facts)
        - Not updating the article as the story develops
        - Missing NewsArticle schema (loses Google News and Discover eligibility)
        - No author byline (news articles without attribution rank poorly)
        - Rewriting the press release without adding original analysis
        
        ## Anti-AI Focus
        News and trend analysis articles are inherently resistant to AI generation because they require real-time information and original reporting. However, as AI tools become faster, the following practices strengthen authenticity:
        
        - **Include original reporting.** Quotes from industry contacts, firsthand observations, or reactions gathered through your own outreach cannot be replicated by AI.
        - **Provide immediate impact analysis from experience.** "Based on what we saw when a similar regulation passed in 2024, we expect..." demonstrates pattern recognition that only comes from domain experience.
        - **Reference specific timeline details.** "The announcement came at 14:00 CET on March 18" or "we first noticed the change in our dashboard on Tuesday morning" -- these granular temporal details signal real-time awareness.
        - **Update visibly with timestamps.** Show when each update was added and what new information it contains. AI-generated news is typically a single static output.
        - **Add your editorial position.** State what you think this means for your audience and why. Taking a stance based on expertise is something AI avoids by design.
        
        ## Example Topics by Niche
        - Energy: "Energieplafond 2026: dit zijn de nieuwe tarieven en regels"
        - Telecom: "5G uitrol Nederland: welke gebieden krijgen dekking in 2026?"
        - SaaS: "Google algorithm update March 2026: what changed and what to do"
        - E-commerce: "Nieuwe EU-regels voor webshops: dit moet je weten over de Digital Services Act"
        - Local services: "Nieuwe Wet kwaliteitsborging bouw: impact voor aannemers en opdrachtgevers"
        
      • pillar-page.md 5.5 KB
        # Technique 20: Ultimate Guide / Pillar Pages
        
        ## What It Is
        Long-form content that covers an entire topic in depth and acts as the central hub of a topic cluster. Captures broad head terms and establishes topical authority.
        
        ## When to Use
        Informational/navigational intent -- user wants a complete overview of a broad topic. Use when building topic clusters. The pillar page links to and from all supporting content (how-tos, definitions, comparisons, FAQs).
        
        ## Structure Template
        
        ```
        H1: "The Complete Guide to [Topic] [Year]"
            (e.g., "Alles over zonnepanelen in 2026: de complete gids")
        
        H2: Table of contents
            Clickable anchor links to all major sections
            Acts as a mini-sitemap for the topic
        
        H2: What is [topic]? (Quick overview)
            2-3 paragraphs establishing context
            Link to dedicated definition article
        
        H2: [Core subtopic 1]
            H3: Key aspect A
            H3: Key aspect B
            Link to dedicated cluster article for deeper detail
        
        H2: [Core subtopic 2]
            H3: Key aspect A
            H3: Key aspect B
            Link to dedicated cluster article
        
        H2: [Core subtopic 3-6]
            Same pattern: overview + link out to cluster content
        
        H2: How to get started / Practical steps
            Actionable summary of the process
            Links to relevant how-to articles
        
        H2: Costs and pricing
            Overview of costs, budget ranges
            Link to dedicated pricing/comparison content
        
        H2: Common mistakes
            Top 5-7 mistakes with brief explanations
        
        H2: FAQ
            5-10 questions covering breadth of the topic
            Each answer 2-4 sentences with link to deeper content
        
        H2: Resources and next steps
            Curated internal links organized by user intent
        ```
        
        ## Word Count
        3,000-5,000+ words (covers the whole topic without padding -- every section earns its place)
        
        ## Schema Markup
        - **Primary:** Article (with headline, author, datePublished, dateModified)
        - **Secondary:** FAQ (for the FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** Paragraph snippet for the overview, list snippets for subtopic sections
        - **Target:** The "What is" section for definition snippets, ToC for list snippets
        - **Tip:** Each H2 section should have a self-contained opening paragraph that can stand alone as a snippet
        
        ## CTA Placement
        - After table of contents: free resource, checklist, or tool
        - Mid-article (every 800-1,000 words): contextual CTAs matching section topic
        - End of article: primary conversion CTA (consultation, trial, comparison tool)
        - Sticky sidebar CTA on desktop
        
        ## Internal Linking Strategy
        - **Hub model:** Link OUT to every cluster article (how-tos, definitions, comparisons, FAQs)
        - **Receive links:** Every cluster article links back to this pillar page
        - **Cross-pillar:** Link to related pillar pages for adjacent topic clusters
        - **Breadcrumb:** Pillar page sits one level below the category/topic page
        - **Anchor text variety:** Use descriptive anchor text, not "click here" or "read more"
        
        ## Key Success Factors
        1. **Table of contents is essential:** Users and Google both need to navigate 3,000+ word pages
        2. **Overview, not exhaustive:** Cover each subtopic enough to be useful, then link to the deep-dive
        3. **Regular updates:** Pillar pages must be updated at least quarterly to maintain rankings
        4. **Strong internal linking:** Every section links to at least one cluster article
        5. **Visual breaks:** Use images, tables, callout boxes, and diagrams every 300-500 words
        6. **Scannable structure:** Users jump between sections -- each H2 must work independently
        7. **E-E-A-T signals:** Author byline, publication date, update date, methodology notes
        
        ## Common Mistakes
        - Writing 5,000 words of shallow content instead of 3,000 words of substantive content
        - Not linking to cluster articles (defeats the purpose of a pillar page)
        - No table of contents (readers bounce from long pages without navigation)
        - Treating it as a one-time publish (pillar pages need regular updates)
        - Duplicating cluster content instead of summarizing and linking
        - Missing the FAQ section (loses long-tail keyword coverage)
        - No visual elements (wall-of-text pillar pages have high bounce rates)
        
        ## Anti-AI Focus
        Pillar pages are less commonly AI-generated wholesale due to their length and complexity, but individual sections can still read as generic. To ensure authenticity:
        
        - **Inject editorial perspective throughout.** Each section overview should reflect a point of view or prioritization that comes from real expertise, not just neutral summarization.
        - **Reference your own cluster content by name.** Linking to your specific articles (with context about what readers will find there) creates a web of evidence that a single AI prompt cannot produce.
        - **Include market-specific data and timelines.** Reference current Dutch pricing, regulation changes, or market conditions with specific dates and figures.
        - **Add "what we have seen" observations.** Brief editorial notes about trends, common misconceptions, or shifts you have noticed in your industry ground the content in lived experience.
        - **Update visibly.** Show a changelog or "last updated" note with specific details about what changed and why. AI-generated content is typically published once and never revisited.
        
        ## Example Topics by Niche
        - Energy: "Alles over zonnepanelen: kosten, subsidie, terugverdientijd en installatie"
        - Telecom: "Internet in Nederland: alles over providers, snelheden en abonnementen"
        - SaaS: "The complete guide to marketing automation in 2026"
        - E-commerce: "Webshop starten: de complete gids van idee tot eerste verkoop"
        - Local services: "Verbouwen in Nederland: vergunningen, kosten, aannemers en subsidies"
        
      • product-page.md 4.5 KB
        # Technique 31: E-Commerce Product Pages
        
        ## What It Is
        Commercial/transactional content for a single product, combining persuasive copywriting with technical specifications and trust signals. The final destination before purchase. Captures "[product name]", "[product] kopen", and "[product] + [specification]" searches.
        
        ## When to Use
        Transactional intent -- user wants to buy a specific product or is evaluating it before adding to cart. Every product in your catalog needs a unique, optimized product page.
        
        ## Structure Template
        
        ```
        H1: "[Product Name] -- [Key Differentiator]"
            (e.g., "SolarEdge 370Wp Zonnepaneel -- Beste rendement per m2")
        
        Product hero section:
            High-quality images (multiple angles, in-use, detail shots)
            Price (with any discounts visible)
            Add to cart / Request quote button
            Key specs: 3-5 bullet points
            Trust signals: rating stars, review count, delivery info
        
        H2: Product description
            2-3 paragraphs: what it is, who it's for, key benefit
            First-person or experience-based language where possible
        
        H2: Key features and specifications
            Specs table: dimensions, weight, performance, compatibility
            Feature explanations: what each spec means for the user
        
        H2: What's included
            Complete list of what comes in the box/package
        
        H2: Who is this for?
            2-3 use cases or customer profiles
            "Perfect for..." / "Not ideal if..."
        
        H2: Customer reviews
            Aggregated rating + individual reviews
            Photo reviews if available
        
        H2: FAQ
            3-5 product-specific questions
            Shipping, returns, compatibility, installation
        
        H2: Related products
            3-6 complementary or alternative products
        ```
        
        ## Word Count
        1,000-1,500 words (concise and scannable -- every word must serve a purpose)
        
        ## Schema Markup
        - **Primary:** Product (name, description, image, sku, brand)
        - **Secondary:** Offer (price, priceCurrency, availability, url)
        - **Tertiary:** AggregateRating + Review (if reviews are present)
        
        ## Featured Snippet Strategy
        - **Format:** Product rich result (not a traditional snippet)
        - **Target:** Product schema with price, availability, and rating for rich results in search
        - **Tip:** Ensure price, currency, and availability are always accurate and up to date
        
        ## CTA Placement
        - **Primary:** Add to cart button visible without scrolling (above the fold)
        - **Secondary:** Sticky add-to-cart bar on scroll
        - End of description: secondary CTA ("request a quote", "compare with similar")
        - After reviews: reinforcement CTA
        
        ## Internal Linking Strategy
        - Link to category page (breadcrumb navigation)
        - Link to related/complementary products
        - Link to buying guide for the product category
        - Link to how-to content for installation or usage
        - Receive links from category pages, buying guides, and roundups
        
        ## Anti-AI Focus
        Product pages benefit from specific, authentic customer language. Include real testimonials with names and details. Manufacturer-spec rewrites are easy for AI to generate, but genuine customer experiences, specific use-case observations, and honest "not ideal if..." assessments are much harder to fake. Prioritize first-hand product knowledge and real-world context over polished but generic descriptions.
        
        ## Key Success Factors
        1. **Unique descriptions:** Never use manufacturer copy -- write original content for every product
        2. **Benefits over features:** "Produces 15% more energy per m2" not just "370Wp output"
        3. **High-quality images:** Multiple angles, lifestyle shots, and zoom capability
        4. **Social proof above the fold:** Star rating and review count visible immediately
        5. **Complete specs table:** Technical buyers need detailed specifications
        6. **Clear pricing:** No hidden costs -- show total price including applicable taxes
        
        ## Common Mistakes
        - Copy-pasting manufacturer descriptions (duplicate content, no differentiation)
        - Single product image (multiple images increase conversion significantly)
        - No customer reviews or ratings on the page
        - Hiding the price or making it hard to find
        - Missing Product+Offer schema (loses rich result eligibility)
        - No "who is this for" context (specs without context confuse non-technical buyers)
        - Thin content (50-word descriptions that add no value)
        
        ## Example Topics by Niche
        - Energy: "SolarEdge 370Wp zonnepaneel kopen -- specificaties, prijs en installatie"
        - Telecom: "KPN Experia Box v10a -- specificaties en compatibiliteit"
        - SaaS: "Acme SaaS Pro Plan -- features, pricing and what's included"
        - E-commerce: "Samsung Galaxy S26 kopen -- prijs, specs en reviews"
        - Local services: "Bosch warmtepomp Compress 7000i AW -- specificaties en installatie"
        
      • programmatic-page.md 4.7 KB
        # Technique 40: Programmatic SEO Pages
        
        ## What It Is
        Template-driven pages generated at scale from structured data, targeting long-tail keyword patterns. Each page targets a specific variation of a repeating search pattern (e.g., "[service] in [city]", "[tool] for [industry]", "[metric] by [country]"). Quality gates are essential to avoid thin content penalties.
        
        ## When to Use
        Informational/commercial intent -- user searches for a specific variation of a common pattern. Use when you have structured data that can populate hundreds or thousands of unique, genuinely useful pages. Only viable when each page provides real value beyond what a single page with filters could offer.
        
        ## Structure Template
        
        ```
        H1: "[Dynamic Variable A] [Static Connector] [Dynamic Variable B]"
            (e.g., "Gemiddelde energieprijs in [Stad] -- [Jaar]")
        
        Data summary section:
            Key metric or answer prominently displayed
            Comparison to national average or benchmark
            Last updated date
        
        H2: [Variable A] in [Variable B]: overzicht
            3-5 sentences of template text with dynamic data points
            Must read naturally -- not obviously template-generated
        
        H2: Key data / Statistics
            Table or data visualization with location/variable-specific data
            Sourced from your structured dataset
        
        H2: How [Variable B] compares
            Comparison to similar entities (nearby cities, similar industries)
            Data table or chart
        
        H2: Trends over time
            Historical data if available
            Chart showing change over time for this specific variable
        
        H2: What this means for you
            Actionable interpretation of the data
            Template text with conditional logic based on data values
        
        H2: Related pages
            Links to similar programmatic pages (nearby cities, related topics)
            Links to parent hub page
        
        H2: Methodology
            Brief explanation of data source and update frequency
            Link to detailed methodology page
        ```
        
        ## Word Count
        500-1,200 words per page (quality over quantity -- thin pages at scale create more problems than they solve)
        
        ## Schema Markup
        - **Primary:** Depends on content type (Dataset, LocalBusiness, Product, Event)
        - **Secondary:** BreadcrumbList (essential for large-scale site architecture)
        
        ## Featured Snippet Strategy
        - **Format:** Table snippet for data-driven queries
        - **Target:** The key data table or comparison table
        - **Tip:** Structured data makes programmatic pages highly eligible for rich results
        
        ## CTA Placement
        - After data summary: relevant tool, calculator, or comparison
        - After "what this means": service page or conversion CTA
        - End of page: parent category or related pages for continued browsing
        
        ## Internal Linking Strategy
        - **Hub-and-spoke:** All programmatic pages link to a parent hub page
        - **Sibling links:** Related pages link to each other (nearby cities, related categories)
        - **Upward:** Link to pillar content for the broader topic
        - **Receive links:** Hub page and blog content link to programmatic pages
        - **Breadcrumbs:** Essential for helping search engines understand the hierarchy
        
        ## Key Success Factors
        1. **Quality gates are essential:** Every page must pass a minimum quality threshold before indexing
        2. **Unique value per page:** Each page must contain data or insight not available on other pages
        3. **Natural language:** Template text must read naturally -- not like a mad-libs fill-in
        4. **Conditional logic:** Different data values should trigger different text (not the same copy for all)
        5. **Noindex thin pages:** Pages with insufficient data should be noindexed rather than published
        6. **Internal linking structure:** Without proper hub-and-spoke linking, Google will not crawl thousands of pages
        7. **Monitor for quality:** Regularly audit a sample of programmatic pages for quality degradation
        
        ## Common Mistakes
        - **No quality gates:** Publishing thousands of thin, nearly-identical pages (Google penalty risk)
        - Template text that reads obviously generated ("Welcome to our page about [city]")
        - No conditional logic (same text regardless of whether data is high, low, or missing)
        - Publishing pages with missing or insufficient data
        - No internal linking strategy (orphaned programmatic pages will not get indexed)
        - Not monitoring page quality over time (data changes can create broken or misleading pages)
        - Scaling before validating (always test with 50-100 pages before generating thousands)
        
        ## Example Topics by Niche
        - Energy: "Energieprijzen per gemeente -- [Stad]: gemiddeld tarief en vergelijking"
        - Telecom: "Glasvezel beschikbaarheid [Postcode] -- providers en snelheden"
        - SaaS: "[Tool] for [Industry]: features, pricing, and alternatives"
        - E-commerce: "[Product categorie] prijzen in [Land] -- actueel overzicht"
        - Local services: "[Vakman] in [Stad]: gemiddelde tarieven en beschikbare vakmensen"
        
      • service-page.md 4.7 KB
        # Technique 36: Service Pages
        
        ## What It Is
        Commercial/transactional content describing a specific service offering, its benefits, process, and pricing. The core conversion page for service-based businesses. Captures "[service]", "[service] inhuren", "[service] in [location]", and "[service] kosten" searches.
        
        ## When to Use
        Commercial/transactional intent -- user is evaluating a specific service or ready to inquire. Every distinct service offering needs its own page. Do not combine multiple services on one page.
        
        ## Structure Template
        
        ```
        H1: "[Service Name] -- [Key Benefit or Location]"
            (e.g., "SEO Advies voor MKB -- Meer organisch verkeer binnen 3 maanden")
        
        Hero section:
            H1 + 2-sentence value proposition
            Primary CTA: "Vraag een offerte aan" or "Plan een gesprek"
            Trust signals: years experience, number of clients, rating
        
        H2: What we do
            Clear description of the service (3-4 sentences)
            Key deliverables as bullet points
        
        H2: Who this is for
            3-4 ideal client profiles
            "This service is for you if..."
        
        H2: How it works (process)
            H3: Step 1: [Phase name]
                What happens, what you deliver, timeline
            H3: Step 2: [Phase name]
            H3: Step 3: [Phase name]
            H3: Step 4: [Delivery/Results]
        
        H2: What you get (deliverables)
            Specific list of deliverables and outcomes
            Tangible where possible: "monthly report", "content calendar", "technical audit"
        
        H2: Results we've achieved
            2-3 client results with specific numbers
            Link to full case studies
        
        H2: Pricing / Investment
            Starting prices or price ranges
            What affects the price
            Link to detailed pricing page if applicable
        
        H2: Why choose us
            3-5 differentiators (specific, not generic)
            Credentials, certifications, experience
        
        H2: FAQ
            5-7 service-specific questions
            Process, timeline, pricing, guarantees
        
        H2: Ready to get started?
            Final CTA with contact form or booking link
        ```
        
        ## Word Count
        1,200-2,000 words (thorough enough to answer all questions, concise enough to keep focus on conversion)
        
        ## Schema Markup
        - **Primary:** Service (name, description, provider, serviceType)
        - **Secondary:** LocalBusiness (if location-specific service)
        - **Tertiary:** FAQ (for the FAQ section)
        
        ## Featured Snippet Strategy
        - **Format:** List snippet for "what does [service] include" queries
        - **Target:** The deliverables section or the process steps
        - **Alternative:** Paragraph snippet for "[service] meaning/definition" queries
        
        ## CTA Placement
        - **Hero section:** Primary CTA above the fold
        - **After process:** "Ready to start?" CTA
        - **After results:** Social proof reinforcement CTA (strongest conversion point)
        - **Sticky:** Persistent CTA button or bar on scroll
        - **End of page:** Full contact form or booking widget
        
        ## Internal Linking Strategy
        - Link to case studies for proof of results
        - Link to related services (cross-sell opportunities)
        - Link to blog content that supports the service's value proposition
        - Link to pricing page for detailed pricing
        - Link to about page for company credibility
        - Receive links from homepage, navigation, blog content, and location pages
        
        ## Key Success Factors
        1. **One service per page:** Do not combine SEO and SEA on one page -- separate pages rank better
        2. **Clear process:** Users want to know exactly what happens after they contact you
        3. **Specific results:** "37% more organic traffic in 4 months" not "we get you more traffic"
        4. **Pricing transparency:** At minimum, provide starting prices or ranges -- hidden pricing loses leads
        5. **Social proof with numbers:** Client results, years of experience, number of projects completed
        6. **Strong above-the-fold CTA:** The inquiry form or booking button must be immediately visible
        7. **FAQ addresses objections:** Use the FAQ to handle "is it worth it" and "how long does it take" questions
        
        ## Common Mistakes
        - Combining multiple services on one page (dilutes rankings and confuses visitors)
        - No pricing information at all (users leave to find a competitor who shows pricing)
        - Generic differentiators ("we care about quality" -- everyone says this)
        - No case studies or results (claims without evidence do not convince)
        - Missing the process section (users fear the unknown)
        - No clear CTA or the CTA is buried below the fold
        - Writing about what you do instead of what the client gets
        
        ## Example Topics by Niche
        - Energy: "Zonnepanelen installatie -- professionele montage met 10 jaar garantie"
        - Telecom: "Zakelijk internet aanvragen -- glasvezel voor bedrijven"
        - SaaS: "SEO audit service -- complete technische en content analyse van je website"
        - E-commerce: "Webshop laten bouwen -- van ontwerp tot lancering in 6 weken"
        - Local services: "Badkamer renovatie Amsterdam -- van ontwerp tot oplevering"
        
      • statistics-page.md 5.2 KB
        # Technique 22: Statistics / Data Pages
        
        ## What It Is
        Data-driven content that curates, visualizes, and contextualizes statistics around a specific topic. Designed as a link magnet -- journalists, bloggers, and researchers link to these pages as primary sources.
        
        ## When to Use
        Informational intent -- user is researching data to support content, presentations, or decisions. Captures "[topic] statistics", "[topic] data [year]", and "how many [topic] in [country]" searches. Highest link-earning content type in SEO.
        
        ## Structure Template
        
        ```
        H1: "[Topic] Statistics [Year]: [X] Key Facts and Trends"
            (e.g., "Zonnepanelen statistieken 2026: 40 feiten en cijfers over Nederland")
        
        H2: Key highlights
            5-7 most compelling statistics as a bulleted list
            Each stat includes source attribution
            Optimized for list featured snippet
        
        H2: [Category 1] statistics
            H3: [Specific stat area]
                Stat + context + source
                Visual: chart, graph, or infographic
            H3: [Specific stat area]
        
        H2: [Category 2] statistics
            Same pattern with data visualizations
        
        H2: [Category 3: Regional/Dutch data]
            Netherlands and European-specific statistics
            Compare to global averages where relevant
        
        H2: Trends and analysis
            What the data means -- expert interpretation
            Year-over-year comparisons
            Forward-looking projections with caveats
        
        H2: Methodology and sources
            How data was collected/curated
            Full source list with links
            Last updated date
        
        H2: FAQ
            3-5 data-related questions
        ```
        
        ## Word Count
        2,000-3,500 words (data-dense -- every sentence should contain or contextualize a number)
        
        ## Schema Markup
        - **Primary:** Article (with dateModified prominently displayed)
        - **Secondary:** Dataset (name, description, distribution, temporalCoverage)
        
        ## Featured Snippet Strategy
        - **Format:** List snippet (statistics queries strongly trigger numbered/bulleted lists)
        - **Target:** The "Key highlights" bulleted list at the top
        - **Alternative:** Table snippet for comparison statistics
        
        ## CTA Placement
        - After key highlights: downloadable PDF or infographic
        - Mid-article: related tool or calculator that uses the data
        - End of article: newsletter signup for data updates, or related analysis content
        
        ## Internal Linking Strategy
        - Link to definition articles for terms behind the statistics
        - Link to analysis/thought leadership pieces that interpret the data
        - Link to how-to articles for acting on the data
        - Link to comparison pages for product/service statistics
        - Receive links from blog posts, pillar pages, and external sites (link magnet)
        
        ## Key Success Factors
        1. **Source everything:** Every single statistic must have a linked source citation
        2. **Recency signals:** Display "Last updated: [date]" prominently and update quarterly
        3. **Original analysis:** Do not just list stats -- add interpretation and year-over-year trends
        4. **Visualize the data:** Charts, graphs, and infographics make the page shareable and linkable
        5. **Lead with the most compelling stat:** The most surprising or impactful number goes first
        6. **Dutch/European focus:** Include Netherlands-specific data -- most competitors only cover US data
        7. **Embeddable assets:** Provide charts that other sites can embed with attribution
        
        ## Common Mistakes
        - Statistics without sources (destroys credibility)
        - No update date (readers cannot assess recency)
        - Just listing numbers without context or analysis
        - Using only US/global data with no Dutch or European perspective
        - No visualizations (text-only statistics pages underperform)
        - Not updating annually (stale stats pages lose rankings fast)
        - Too many statistics without categorization (overwhelming wall of numbers)
        
        ## Anti-AI Focus
        Statistics pages have a natural advantage against AI imitation because they depend on verifiable, sourced data. However, AI can still generate plausible-sounding but fabricated statistics. To strengthen authenticity:
        
        - **Link every statistic to its primary source.** AI-generated statistics pages often cite vague sources or none at all. Direct links to CBS, Eurostat, or industry reports are difficult to fake convincingly.
        - **Include your own original data or analysis.** If you have proprietary data (survey results, platform usage data, customer trends), present it alongside third-party statistics. Original data is the strongest anti-AI signal.
        - **Show calculation methodology.** When you derive a number (e.g., "average savings of X per household"), show how you arrived at it. AI cannot replicate transparent methodology.
        - **Add year-over-year context from personal observation.** "This is up 15% from 2025, which aligns with what we have seen in our customer base" combines data with experience.
        - **Date-stamp aggressively.** Include specific collection dates and note when individual statistics were last verified. Stale or undated statistics are a hallmark of AI-generated content.
        
        ## Example Topics by Niche
        - Energy: "Energieprijzen Nederland 2026: 35 statistieken en trends"
        - Telecom: "Internet in Nederland: 30 statistieken over snelheid, dekking en gebruik"
        - SaaS: "Email marketing statistics 2026: 50 benchmarks you need to know"
        - E-commerce: "E-commerce in Nederland: 40 cijfers over online winkelen"
        - Local services: "Woningmarkt statistieken 2026: huizenprijzen per regio"
        
    • anti-slop-ruleset.md 14 KB
      # Technique 02: Anti-AI-Slop Defense System
      
      ## What It Is
      
      A multi-layered detection and rewriting system that identifies and eliminates AI writing patterns from generated content. Based on research from 6+ GitHub anti-slop skills, Wikipedia's WikiProject AI Cleanup (24+ patterns), academic research on semantic ablation, and StyloAI's 31+ stylometric markers. This is not a post-processing "humanizer" -- it is a quality standard built into the writing process from the start.
      
      **Framing note:** Never label this as "humanization" or "AI detection avoidance." Frame it as quality writing standards. The goal is not to trick detection -- it is to produce writing that is genuinely better than default AI output.
      
      ## Why It Works
      
      AI content is detectable because of **semantic ablation** -- RLHF training systematically strips high-entropy information (unique metaphors, specific details, strong opinions), producing text with low burstiness (uniform sentence rhythm) and low perplexity (predictable word choice). These are the exact signals that both human readers and detection algorithms use to identify AI content.
      
      Google's NavBoost system (confirmed in antitrust trial) demotes content that fails to engage users. Bland, detectable AI content drives pogo-sticking (users bouncing back to SERP), which directly hurts rankings.
      
      **Key insight:** Post-processing (running text through a "humanizer") is detectable itself -- it creates a different statistical signature that is distinct from both pure AI and pure human writing. The solution is to generate human-like text from the start, not to fix it after.
      
      ## The Science of Detection
      
      ### Perplexity (How predictable each word is)
      - **AI text**: Low perplexity -- the model picks the statistically most likely next word. "The results clearly demonstrate the significant impact..."
      - **Human text**: Higher perplexity -- humans choose unexpected words, use slang, make idiosyncratic choices. "The numbers surprised us. Not because they were good -- they weren't -- but because they were exactly wrong."
      - **Fix**: Instruct the model to use less common phrasings, domain-specific jargon, and unexpected vocabulary choices
      
      ### Burstiness (Sentence length variation)
      - **AI text**: Remarkably uniform -- 15-25 words per sentence, consistent complexity
      - **Human text**: Wild variation. Like this. And then suddenly a sentence that stretches across multiple clauses, connecting ideas that might not obviously belong together but make sense when you follow the writer's train of thought. Then short again.
      - **Fix**: Explicitly vary sentence length between 3 and 40+ words per section
      
      ### Stylometric Patterns (31+ markers per StyloAI research)
      - Function word frequency, sentence complexity distribution, vocabulary richness, syntactic diversity
      - **AI text**: Clusters tightly in a predictable zone across all markers
      - **Human text**: Much wider distribution with more variation
      
      ## The Tiered Banned Phrase System
      
      ### Tier 1: Always Remove (+3 points each)
      
      These words and phrases are so strongly associated with AI that their presence alone flags content:
      
      **Vocabulary:**
      delve, tapestry, realm, landscape (metaphorical), multifaceted, nuanced, testament, cutting-edge, revolutionary, comprehensive (as adjective), crucial, compelling, vibrant, game-changer, leverage (verb), unlock potential, harness, endeavour, embark, navigate (metaphorical), pivotal, intricate, innovative, seamless, robust, transformative, meticulous, facilitate, utilize, commence, paramount, plethora, myriad, culminate, underscore, bolster, spearhead
      
      **Phrases:**
      - "It's worth noting that..."
      - "In today's [anything] landscape"
      - "Moreover" / "Furthermore" / "Additionally" (as sentence starters)
      - "Let's dive in" / "Without further ado"
      - "In conclusion" / "In summary" / "In essence"
      - "It's important to note that..."
      - "Stands as a testament to..."
      - "Plays a vital/crucial/pivotal role"
      - "In the realm of..."
      - "It goes without saying"
      - "Buckle up" / "Strap in"
      - "The landscape is ever-evolving"
      - "Marking a pivotal moment in the evolution of..."
      
      **Replacements:**
      | AI Word | Human Alternative |
      |---------|------------------|
      | leverage | use |
      | utilize | use |
      | commence | start, begin |
      | facilitate | help, enable |
      | endeavor | try, attempt |
      | robust | strong, solid |
      | seamless | smooth, easy |
      | innovative | new, original |
      | comprehensive | full, complete, thorough |
      | paramount | critical, key |
      | plethora | many, lots of |
      | culminate | end in, lead to |
      | underscore | show, highlight |
      | delve | dig into, look at |
      | testament | proof, sign |
      | landscape | space, market, field |
      | showcase | show |
      
      ### Tier 2: Flag When Clustered (+2 points each, flag at 3+ per article)
      
      These are acceptable in isolation but signal AI when they cluster:
      
      - however (more than 2x per article)
      - here's the thing
      - at the end of the day
      - paired adjectives ("comprehensive and thorough", "innovative and cutting-edge")
      - template openings ("In this article, we'll cover...")
      - firstly / secondly / thirdly
      - moving forward
      - arguably
      - notably
      - essentially
      - that said
      - on the other hand
      - it's no secret that
      
      ### Tier 3: Flag at High Density (+2 points per cluster of 3+)
      
      - Transition words when overused (however, therefore, consequently, meanwhile)
      - Corporate buzzwords (scalable, streamlined, optimize, synergy)
      - Excessive qualifiers (quite, rather, somewhat, fairly, relatively)
      - Hedging language (perhaps, might, could potentially, to some extent)
      
      ## Structural Pattern Detection
      
      ### Patterns to Flag and Rewrite
      
      1. **Uniform sentence length** (+4 points): Three or more consecutive sentences within 3 words of each other in length. Fix: vary dramatically -- mix 5-word punches with 30-word complexes.
      
      2. **Rule-of-three addiction** (+3 points): Habitual grouping of three examples, adjectives, or list items. Fix: use 2, 4, or 5 items instead.
      
      3. **Topic-sentence-support pattern** (+3 points): Every paragraph starting with a topic sentence followed by supporting detail. Fix: start some paragraphs with an example, a question, or a statement that only makes sense after reading on.
      
      4. **Binary contrast pattern** (+3 points): "It's not X, it's Y" or "No X, no Y, just Z." Fix: make the actual argument without the rhetorical frame.
      
      5. **Staccato fragments** (+3 points): "Short. Punchy. Exhausting." -- dramatic fragmentation that feels manufactured. Fix: use fragments sparingly, for genuine emphasis only.
      
      6. **Compulsive summarizing** (+2 points): Summary paragraph at the end of every section. Fix: only summarize if the section is genuinely complex (3+ subsections).
      
      7. **Copula avoidance** (+2 points): "serves as" instead of "is", "functions as" instead of "is". Fix: just say "is" when that's what you mean.
      
      8. **Synonym cycling** (+2 points): Using different words for the same concept in adjacent sentences to appear varied. Fix: repeat the same word if clarity demands it.
      
      9. **Participial tack-ons** (+2 points): "...highlighting the importance of X" or "...underscoring the need for Y" at the end of sentences. Fix: delete or make a separate sentence.
      
      10. **Em dash overuse** (+1 point per em dash beyond 1 per 500 words): AI uses em dashes far more frequently than humans. Fix: use commas, parentheses, or separate sentences.
      
      ### Additional Structural Patterns
      
      11. **The hourglass structure**: AI structures text as synthesis -> detail -> synthesis. Humans start with a specific story, zoom out, dive into details in uneven bursts, and end with something specific again. **Break the hourglass.**
      
      12. **Asymmetric coverage absence**: AI distributes word count evenly across all subtopics. Humans spend 500 words on the interesting part and 50 on the boring-but-necessary part. If every section is roughly the same length, flag it.
      
      ## The Two-Pass Audit System
      
      ### Pass 1: Detection and Rewrite
      
      Scan the complete text against all three tiers plus structural patterns. For each detection:
      1. Quote the flagged text
      2. Explain why it's flagged (tier + specific pattern)
      3. Provide a rewrite that preserves meaning while eliminating the pattern
      4. Apply the rewrite
      
      ### Pass 2: Surviving Pattern Check
      
      After Pass 1 rewrites, re-scan the full text. First-pass rewrites sometimes introduce new AI patterns (the model's natural tendency reasserts). Specifically check for:
      - Recycled transitions that survived the first edit
      - Lingering vocabulary inflation (replaced one AI word with another)
      - Copula swaps that crept back in
      - New structural uniformity introduced by the rewrites
      
      ### The Horoscope Test (Final Check)
      
      For each paragraph, ask: "Could anyone have written this, for anyone, about anything?" If the answer is yes, the paragraph fails regardless of passing the phrase checks. Every paragraph must contain at least one element that is specific to THIS topic, THIS audience, THIS moment.
      
      Score: +5 points for each paragraph failing the Horoscope Test.
      
      ## Risk Scoring
      
      | Score | Risk Level | Action |
      |-------|-----------|--------|
      | 0-5 | Low | Content passes |
      | 6-12 | Medium | Flag specific issues for human review |
      | 13-20 | High | Auto-revise flagged sections |
      | 21+ | Critical | Full rewrite required |
      
      ## Content Patterns to Eliminate
      
      Beyond vocabulary and structure, eliminate these content-level patterns:
      
      1. **No formulaic challenges**: "Despite facing challenges, the company continues to thrive" -- name the actual challenge or cut it
      2. **No vague attributions**: "Studies show..." -> "A 2024 Ahrefs study of 100K pages found..."
      3. **No significance inflation**: "Marking a pivotal moment in the evolution of..." -> Cut the entire phrase. Start with what actually happened.
      4. **Use contractions**: "doesn't" not "does not." "We've found" not "We have found."
      5. **Use colloquial transitions**: "Thing is," "Here's the catch," "The weird part?" instead of "Furthermore," "Moreover," "Additionally"
      6. **Repeat the clearest word** instead of cycling synonyms: If the subject is "customers," don't alternate between "clients," "users," "patrons," and "consumers"
      
      ## Tips
      
      - **The "bar conversation" test**: Would you say this sentence to a colleague at a bar? If not, it's too formal or AI-sounding. "The implementation demonstrated significant improvements in key performance indicators" -> "The thing actually worked. Revenue went up 18%."
      
      - **The Horoscope Test**: Could this paragraph appear in any article, about any topic, for any audience? If yes, it has zero specificity. Every paragraph needs at least one element unique to THIS topic.
      
      - **Self-correction is human**: "At first I thought this was a branding problem -- turns out it was pricing all along" shows the writer's actual thought process. AI does not self-correct mid-argument.
      
      - **Tangential value**: Brief tangents that add context but aren't strictly necessary: "Side note: we tested this during Black Friday, which might have skewed results upward." AI stays rigidly on-topic. Humans digress when the digression adds value.
      
      - **Asymmetric coverage**: Spend 500 words on the interesting part and 50 on the boring-but-necessary part. AI distributes word count evenly. Humans allocate attention where it matters.
      
      - **Imperfections are features**: One slightly awkward phrasing, one mild redundancy, one sentence that could be tighter -- these are signals of human writing. Perfect prose = AI prose.
      
      - **Domain convention awareness**: Academic writing IS more formal than blog writing. Match the expected register for the content type. Don't make an academic paper sound like a blog post, and don't make a blog post sound like a dissertation.
      
      - **The rhythm check**: Count sentence lengths in each paragraph. If standard deviation is < 5 words, rewrite for more variation.
      
      ## Common Mistakes
      
      1. **Thinking "more editing" = more human** -- Over-humanizing creates its own detectable pattern. Don't make every sentence a fragment or start every paragraph with "Look,". One technique used excessively becomes its own signature.
      
      2. **Using "humanizer" tools** -- These are detectable as post-processed text. The statistical signature is different from both pure AI and pure human writing. They solve the wrong problem.
      
      3. **Thinking typos = human** -- Intentional typos are a different pattern than natural typos. Don't add them deliberately.
      
      4. **Ignoring domain conventions** -- Academic writing IS more formal than blog writing. A technical whitepaper should not read like a casual blog post. Match the expected register for the content type.
      
      5. **Removing patterns without adding personality** -- Anti-slop auditing alone produces sterile text. Voice injection (Technique 03) must happen BEFORE anti-slop auditing.
      
      6. **Thinking "different words" = fixed** -- Replacing one AI word with another AI word (e.g., "leverage" -> "harness") accomplishes nothing. Replace with genuinely plain language ("use").
      
      ## Implementation Notes
      
      - The tiered system allows the writing phase to use Tier 2/3 words occasionally (natural) while the audit phase catches clusters
      - Pass 2 is essential -- research shows first-pass rewrites retain 15-20% of AI patterns
      - The Horoscope Test catches generic content that avoids banned words but remains soulless
      - Voice injection (Technique 03) must happen BEFORE anti-slop auditing -- removing patterns without adding personality produces sterile text
      - Anti-detection instructions should be included IN the writing prompt, not as a separate post-processing step
      
      ## Key Sources
      
      - [blader/humanizer](https://github.com/blader/humanizer) -- 1,600+ stars, based on Wikipedia's 24 patterns
      - [conorbronsdon/avoid-ai-writing](https://github.com/conorbronsdon/avoid-ai-writing) -- 102-entry tiered replacement table
      - [hardikpandya/stop-slop](https://github.com/hardikpandya/stop-slop) -- 8 rules, 50-point scoring
      - [aplaceforallmystuff/the-antislop](https://github.com/aplaceforallmystuff/the-antislop) -- Horoscope Test
      - [DonAldente-AI/anti-slop-system](https://github.com/DonAldente-AI/anti-slop-system) -- Naive drafter + ruthless critic pattern
      - [Wikipedia: Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing)
      - [Semantic Ablation (The Register)](https://www.theregister.com/2026/02/16/semantic_ablation_ai_writing/)
      
    • content-freshness.md 4.6 KB
      # Technique 12: Content Freshness & Strategic Updates
      
      ## What It Is
      Systematically updating existing content to maintain and improve rankings -- because Google's freshness signals (confirmed via `bylineDateFreshness` in the API leak) penalize stale content and reward regularly updated pages.
      
      ## Why It Works
      The 2024 Google API leak revealed `bylineDateFreshness` as a ranking signal -- Google evaluates content freshness based on publication and modification dates. For queries with freshness intent (anything with a year, "best," "latest," or trending topics), outdated content drops steadily.
      
      More importantly, updating content is often MORE effective than creating new content:
      - Existing pages already have authority (backlinks, indexed history)
      - Updates trigger re-crawling and re-evaluation
      - Fresh data in an established page can outrank new pages from competitors
      
      **Data point:** Pages updated with substantial new content saw an average 106% increase in organic traffic (HubSpot, 2024 content refresh study).
      
      ## Step-by-Step Process
      
      ### Step 1: Identify Pages to Update
      1. Use page-level SEO data and opportunity detection from your SEO data source
      2. Priority targets:
         - Pages that rank positions 4-20 (close to page 1 -- a refresh could push them up)
         - Pages with declining traffic (were ranking well, now dropping)
         - Pages with outdated data (old statistics, discontinued tools, old years)
         - Pages targeting keywords with freshness intent ("2025" -> needs "2026" update)
      3. Skip pages ranking #1-3 unless competitors are publishing fresher content
      
      ### Step 2: Competitive Freshness Audit
      4. For each target page, check what competitors published recently
      5. If competitors have 2026 data and your page has 2024 data, update urgently
      6. If competitors are adding new sections you don't cover, add them
      7. Use competitor data to identify who's publishing new competing content
      
      ### Step 3: Content Update Depth
      8. **Surface update** (changes < 20%): Update year references, refresh statistics, fix broken links
      9. **Moderate update** (changes 20-50%): Add new sections, update examples, refresh data throughout
      10. **Major overhaul** (changes > 50%): Effectively rewrite the article with new information, keeping the URL and historical authority
      11. Choose depth based on: how far rankings have dropped, how much new information exists, competitor content freshness
      
      ### Step 4: Strategic Updates
      12. Update the publication date ONLY when making substantial changes (not just fixing a typo)
      13. Add "Last updated: [date]" to the page (transparency + trust signal)
      14. Include new information gain elements (new data, new case studies, updated recommendations)
      15. Remove outdated information (old tools, discontinued services, expired advice)
      16. Update internal links to point to newer related content
      
      ### Step 5: Update Schedule
      17. **Evergreen content**: Review every 6 months, update annually minimum
      18. **Trending/seasonal content**: Update before the season/trend starts (e.g., update "Best X 2026" in November 2025)
      19. **Data-heavy content**: Update whenever new data is available (quarterly for fast-moving industries)
      20. **Performance-triggered**: Update immediately when a page drops >5 positions for its primary keyword
      
      ## Tips
      
      - **The "search your own title" trick**: Google your article's exact title. If competitors have newer articles with the same topic, you need to update.
      - **Add "as of [date]" qualifiers**: For time-sensitive claims, adding "as of March 2026" signals freshness to both Google and users.
      - **Don't change the URL**: Updates should keep the same URL to preserve backlinks and indexing history. Redirects lose 5-15% of link equity.
      - **Content pruning boost**: Removing outdated, low-quality pages from your site can improve the performance of remaining pages (quality signal at site level).
      
      ## Common Mistakes
      
      1. **Changing only the date**: Updating "2025" to "2026" without adding new information is detectable and doesn't improve quality
      2. **Creating a new page instead of updating**: Two pages targeting the same keyword = cannibalization. Update the existing page.
      3. **Over-updating**: Making changes every week signals instability. Update when you have substantial new information.
      4. **Deleting old content**: Pages with backlinks should be updated, not deleted. Even if the content is outdated, the URL has authority.
      
      ## Tools Used
      - page-level SEO data -- current page performance and metrics
      - opportunity detection -- pages ranking 4-20 (update candidates)
      - competitor data -- monitor competitor content freshness
      - keyword cluster data -- check if the topic cluster needs expansion during update
      
    • content-types-overview.md 12.1 KB
      # Content Types Overview
      
      Quick-reference index covering all 23 content-type templates. Use this file to pick the right content type for a SERP, check its target word count, dominant schema, snippet format, and H1/H2 skeleton without loading the full per-type file. Load the individual `content-types/<type>.md` only when you need the depth (process detail, anti-AI focus, common mistakes, example topics).
      
      ## Decision table
      
      | Content type | Dominant intent | Word count | Primary schema | Snippet format | H1 pattern | Key H2s (first 3-4) |
      |---|---|---|---|---|---|---|
      | how-to | Informational | 1,500-3,000 | HowTo + FAQ | Ordered list | "How to [Action] [Qualifier]" | Quick answer / Prerequisites / Step-by-step / Common mistakes |
      | definition | Informational | 1,000-2,000 | DefinedTerm + FAQ | Paragraph | "What Is [Term]? [Subtitle]" | Quick definition / How it works / Types / Benefits |
      | pillar-page | Informational + navigational | 3,000-5,000+ | Article + FAQ | Paragraph (list per section) | "The Complete Guide to [Topic] [Year]" | Table of contents / Overview / Core subtopics / Practical steps |
      | faq-page | Informational | 1,500-2,500 | FAQPage + BreadcrumbList | Paragraph (per Q) | "[Topic] FAQ: Veelgestelde vragen over [topic]" | Intro / Category 1 / Category 2 / Costs |
      | statistics | Informational | 2,000-3,500 | Article + Dataset | List | "[Topic] Statistics [Year]: [X] Key Facts and Trends" | Key highlights / Category stats / Regional data / Trends |
      | news | Informational + navigational | 800-1,500 | NewsArticle + FAQ | Paragraph | "[News event]: [Impact or what it means]" | Key takeaways / What happened / Why it matters / What to expect |
      | thought-leadership | Informational | 1,200-2,500 | Article + Person | Paragraph (low priority) | "[Strong opinion or provocative question]" | Conventional wisdom / Why it's wrong / Supporting arguments / Counterarguments |
      | glossary | Informational | 5,000-15,000 (200-400/term) | DefinedTermSet + BreadcrumbList | Paragraph (per term) | "[Industry] Woordenlijst: [X] Begrippen Uitgelegd" | Intro / A / B / C ... |
      | comparison | Commercial investigation | 1,200-2,500 | FAQ + Product | Table | "[Product A] vs [Product B]: [Outcome promise]" | Quick verdict / At-a-glance table / How we compared / Head-to-head |
      | listicle | Commercial investigation | 2,000-4,000 | ItemList + FAQ | List | "Best [Category] for [Audience/Use Case] in [Year]" | Top picks at a glance / How we evaluated / 1. Top pick / 2. Second pick |
      | alternatives | Commercial investigation | 1,500-3,000 | ItemList + FAQ | List | "[X] Alternatieven: [N] Beste Opties in [Year]" | Why look for alternatives / Comparison table / Alternative 1 / Alternative 2 |
      | product-review | Commercial investigation | 1,500-2,500 | Review + Product | Paragraph | "[Product] Review [Year]: [Honest verdict]" | Quick verdict / What is [Product] / Our experience / Pros |
      | buying-guide | Commercial investigation | 2,000-3,500 | Article + FAQ | List | "[Category] Kopen: Waar Moet Je Op Letten?" | Quick checklist / Criterion 1 / Criterion 2 / Budget guide |
      | product-page | Transactional | 1,000-1,500 | Product + Offer + AggregateRating | Product rich result | "[Product Name] — [Key Differentiator]" | Description / Features and specs / What's included / Who is this for |
      | category-page | Commercial + navigational | 300-1,000 (editorial) | CollectionPage + ItemList | List | "[Category Name]" | Intro / Product grid / Koopgids / Subcategories |
      | landing-page | Transactional | 800-1,200 | WebPage or Service + FAQ | Not primary | "[Clear value proposition in one line]" | Problem statement / Solution / How it works / Social proof |
      | pricing-page | Transactional | 500-1,500 | Product + Offer | Table | "Prijzen" or "[Product] Pricing" | Pricing tiers / Feature comparison table / Which plan / What's included |
      | integration-page | Commercial + navigational | 800-1,500 | SoftwareApplication + HowTo | List | "[Your Product] + [Partner Tool] Integratie" | What it does / Key features / How to set it up / Use cases |
      | service-page | Commercial + transactional | 1,200-2,000 | Service + LocalBusiness + FAQ | List | "[Service Name] — [Key Benefit or Location]" | What we do / Who this is for / How it works / Deliverables |
      | location-page | Local + transactional | 1,200-1,500 | LocalBusiness + Service + FAQ | Paragraph | "[Service] in [Location]" | Local context / Services in [Location] / Why choose us / Recent projects |
      | case-study | Commercial investigation | 1,500-2,500 | Article + Organization | Paragraph | "[Client/Industry]: [Key Result in Numbers]" | Overview / The challenge / Our approach / The results |
      | about-page | Navigational + trust | 500-1,500 | Organization + Person | Knowledge Panel (not snippet) | "Over [Brand]" or "About [Brand]" | Mission / Our story / Our team / Expertise |
      | programmatic-page | Informational + commercial | 500-1,200 | Depends on type + BreadcrumbList | Table | "[Variable A] [Connector] [Variable B]" | Data summary / Key data / How [B] compares / Trends over time |
      
      ## Intent → content type mapping
      
      Search intent is the primary filter. Pick based on what the top 5 SERP results are actually doing, not what you think they should be doing.
      
      **Informational.** User wants to understand something. Not buying yet.
      - "how to [x]" / "[x] tutorial" → how-to
      - "what is [x]" / "[x] meaning" → definition
      - "[topic] explained" / "guide to [topic]" → pillar-page
      - "[topic] questions" / PAA cluster → faq-page
      - "[topic] statistics" / "[topic] data [year]" → statistics
      - "[term] vocabulary" / 30+ terms in niche → glossary
      - "[event] [year]" / breaking topic → news
      - "why [industry belief]" / contrarian query → thought-leadership
      
      **Commercial investigation.** User is shopping but hasn't picked a product.
      - "best [category] for [audience]" → listicle
      - "[A] vs [B]" → comparison
      - "[product] alternatives" → alternatives
      - "[product] review" → product-review
      - "[category] buying guide" / "what to look for in [x]" → buying-guide
      - "[service] results" / "[industry] case study" → case-study
      
      **Transactional.** User is ready to convert.
      - "[product name]" / "[product] kopen" → product-page
      - "[product] pricing" / "[product] kosten" → pricing-page
      - "[service] aanvragen" / "[tool] free trial" → landing-page
      - "[service] inhuren" / "[service] kosten" → service-page
      - "[tool A] + [tool B] integration" → integration-page
      - "[service] in [city]" / "[service] near me" → location-page
      
      **Navigational + trust.** User is verifying who you are.
      - "[brand] about" / "who is behind [brand]" → about-page
      - "[category]" root browse → category-page
      - Long-tail pattern at scale with structured data → programmatic-page
      
      Note: several content types span two intent classes. Pillar pages are informational but also navigational (hub for the cluster). Category pages are commercial but also navigational (people type the category name to browse). Pick the dominant intent, but don't be surprised when a page has to satisfy both.
      
      A few more overlaps worth flagging. Integration pages and category pages both straddle commercial and navigational. Programmatic pages can skew either informational or commercial depending on the variable. Service pages are commercial in SERP but behave like transactional once the visitor lands.
      
      ## When to load the full per-type file
      
      The decision table gives you enough to pick the type and draft the skeleton. Load the full `content-types/<type>.md` when:
      
      - **pillar-page**: you need the hub-and-spoke linking architecture, the visual-break rhythm (image/table every 300-500 words), and the regular-update cadence rules. The overview doesn't capture the "every cluster article links back" rule, which is the whole point of the pattern.
      - **comparison**: you need the full 10-section framework, especially the "verdict first per criterion" rule at the H3 level. Comparison articles fail hard if you just know the H2 skeleton.
      - **how-to**: you need the quick-answer targeting (40-60 words, ordered list after the most relevant H2) and the step-by-step "one action per step" constraint. The anti-AI section on error messages and version numbers is the highest-signal part.
      - **case-study**: you need the PAS-adjacent structure (challenge → approach → results → takeaways) and the "lead with the number in H1 and first 50 words" rule. The overview skips the client-permission and visual-proof requirements.
      - **product-review**: you need the "who should skip" framing and the Google review-update requirements around hands-on evidence.
      - **programmatic-page**: you need the quality-gate logic (noindex thin pages, conditional template copy, sibling linking). Shipping programmatic without reading the full file gets you penalized.
      - **listicle**: you need the "methodology section + updated quarterly" rule and the 7-10 item sweet spot.
      - **glossary**: you need the A-Z navigation pattern, anchor-link scheme, and the cross-reference style between terms.
      - **location-page**: you need the doorway-page warning. Location pages without genuinely unique per-city content get penalized. The overview doesn't make this loud enough.
      
      For everything else the overview table is usually enough to draft the first pass. Go deeper if the draft feels generic or if you're unsure about a specific structural element.
      
      ## Cross-reference: YMYL-sensitive types
      
      YMYL ("your money or your life") topics trigger elevated E-E-A-T and Trust scrutiny from Google's quality raters. The content types that live closest to YMYL need named authors, visible credentials, transparent methodology, and verifiable evidence. That list: **pricing-page**, **service-page**, **product-review** (Google's review update explicitly targets this type), **case-study**, **about-page** (the page raters read to assess the whole site's trust), **thought-leadership**, and any **news** article about health, finance, legal, or safety topics. For these, skip stock photos, add real author bylines with linked credentials, include dated evidence (screenshots, numbers, timelines), and publish visible update logs. If you're writing in a regulated niche (medical, financial advice, legal), the Trust bar is higher than any of the per-type files capture. Add a subject-matter expert reviewer byline and cite primary sources (regulators, peer-reviewed studies) rather than secondary blogs.
      
      ## Word count targets by intent
      
      The overriding rule: match the average word count of the top 5 ranking results, then add roughly 10%. Don't pad to hit a number. Don't under-write to stay "concise" when the SERP is hungry for depth.
      
      | Intent class | Typical range | Rule of thumb |
      |---|---|---|
      | Informational (simple) | 800-1,500 | Match SERP average; news and short definitions fall here |
      | Informational (deep) | 1,500-3,500 | How-to, statistics, FAQ, buying guide: depth signals expertise |
      | Informational (hub) | 3,000-5,000+ | Pillar pages only; every section earns its place |
      | Commercial investigation | 1,200-3,000 | Comparison, listicle, alternatives, review: enough to build confidence before conversion |
      | Transactional | 500-1,500 | Pricing, product, landing, integration: short and scannable, conversion over coverage |
      | Trust / navigational | 500-1,500 | About, category: authentic over long, unique content over padding |
      | Programmatic at scale | 500-1,200 per page | Quality gates first; noindex anything thin |
      
      When the SERP has mixed formats (one pillar, three listicles, one FAQ), write for the dominant format but cover the angles the minority formats introduce. If the top 5 are all 3,500 words and yours is 1,200, you're not competing. You're just hoping.
      
      ## Related reference files
      
      If you need the underlying methodologies rather than the content-type skeletons, load these from the same `references/` directory:
      
      - `intent-matching.md`: the full intent-classification logic and SERP-reading playbook
      - `serp-driven-writing.md`: how to read the top 5 and extract the winning format
      - `structured-data-snippets.md`: the full schema-markup catalog with examples
      - `anti-slop-ruleset.md`: the banned-vocabulary list and structural tells
      - `voice-injection-playbook.md`: how to inject first-person experience into any type
      - `eeat-signal-embedding.md`: where and how to place author, credentials, and evidence
      
    • quality-scoring.md 6.9 KB
      # Technique 16: Quality Scoring System
      
      ## What It Is
      A 100-point composite scoring system that evaluates AI-generated content across five dimensions before output. Content scoring below threshold is automatically revised. This is the final quality gate preventing bad content from being published.
      
      ## Why It Works
      Organizations using structured quality gates report 45% fewer post-publication content issues. The scoring system provides objective, repeatable assessment and identifies WHICH areas need improvement rather than a binary pass/fail.
      
      ## Scoring Breakdown (100 Points Total)
      
      ### Content Quality (30 points)
      
      **Topical Completeness (0-10):**
      - Are all SERP-common topics covered? (from brief's competitor analysis)
      - Are PAA questions from the brief addressed?
      - Scoring: 10 = all topics covered, 7 = 80%+, 5 = 60%+, 3 = 40%+, 0 = below 40%
      
      **Depth of Coverage (0-10):**
      - Does each major section contain specific examples, data points, or analysis?
      - Are claims supported with evidence rather than stated as fact?
      - Scoring: count sections with at least one specific example/data point. 10 = all sections, proportional down
      
      **Originality (0-10):**
      - Does the content contain at least one section adding information not found in the top 10 SERP results?
      - Are there unique insights, original analysis, contrarian perspectives, or proprietary data?
      - Scoring: 10 = multiple unique sections, 7 = one strong unique section, 4 = unique framing of known info, 0 = pure rehash
      
      ### SEO Optimization (25 points)
      
      **Keyword Placement (0-8):**
      - Primary keyword in meta title: +2
      - Primary keyword in H1: +2
      - Primary keyword in first 100 words: +2
      - Primary keyword or variation in 2+ H2s: +2
      
      **Internal Links (0-5):**
      - 3-5 verified internal links per 1,000 words: +3
      - All links validated against real sitemap: +2
      - Deduct 2 points per hallucinated link
      
      **Meta Quality (0-4):**
      - Meta title under 60 chars with keyword + hook: +2
      - Meta description under 155 chars with CTA: +2
      
      **Schema Markup (0-4):**
      - Correct primary schema for content type: +2
      - FAQ schema present (if FAQ section exists): +2
      
      **Featured Snippet Targeting (0-4):**
      - Target snippet format identified and implemented: +2
      - 40-60 word direct answer at correct position: +2
      
      ### E-E-A-T Signals (15 points)
      
      **First-Person Experience (0-5):**
      - Uses "I/we tested," "in my experience," "our data shows": +2
      - Contains specific case study or personal testing data: +3
      - Scoring: presence of genuine experience signals, not just the words
      
      **Specific Data & Examples (0-5):**
      - Named companies, specific numbers, dated events: +2
      - At least one example that is not a common/obvious example: +3
      - Deduct points for "many companies" or "experts agree" without names
      
      **Source Attribution (0-5):**
      - Claims backed by named sources with dates: +3
      - External links to authoritative references: +2
      
      ### Anti-Slop Score (15 points)
      
      **Zero Tier 1 Phrases (0-5):**
      - 5 = zero Tier 1 words/phrases detected
      - Deduct 1 point per Tier 1 detection (minimum 0)
      
      **Burstiness (0-3):**
      - Measure sentence length standard deviation
      - 3 = high variation (SD > 8 words), 2 = moderate (SD 5-8), 1 = low (SD 3-5), 0 = uniform
      
      **Voice Consistency (0-3):**
      - Content matches the loaded voice document characteristics
      - 3 = strong match, 2 = moderate, 1 = weak, 0 = no voice detected
      
      **Horoscope Test (0-4):**
      - Count paragraphs that pass the specificity check (topic-specific, audience-specific)
      - 4 = all pass, 3 = 80%+, 2 = 60%+, 1 = 40%+, 0 = below 40%
      
      ### AI Citation Readiness (15 points)
      
      **Answer-First Formatting (0-5):**
      - Key sections lead with extractable 40-60 word answers: +3
      - Direct answer appears within first 100 words of the article: +2
      
      **FAQ Schema (0-3):**
      - FAQPage schema present with 3+ properly formatted Q&As: +3
      
      **Claim-Evidence Pairs (0-4):**
      - Data claims include named source + date: +2
      - Statistics include specific numbers (not "many" or "most"): +2
      
      **Entity Richness (0-3):**
      - Named entities (companies, people, tools) used instead of generic references: +2
      - Entity relationships explicitly stated: +1
      
      ## Readability Targets
      
      | Audience | Flesch Reading Ease | Grade Level | When to Use |
      |----------|-------------------|-------------|-------------|
      | General consumer (NL) | 60-70 | Grade 7-9 | Comparison sites, how-tos, consumer guides |
      | Professional / B2B | 50-60 | Grade 9-12 | SaaS content, industry analysis |
      | Technical | 40-50 | Grade 12+ | Developer docs, technical guides |
      
      Readability is assessed but NOT part of the 100-point score (since optimal readability varies by audience and content type).
      
      ## Quality Gate Actions
      
      | Score Range | Risk Level | Action |
      |-------------|-----------|--------|
      | 85-100 | Ready | Output for publication |
      | 70-84 | Minor issues | Output with annotations on weak areas for human review |
      | 60-69 | Needs revision | Auto-revise weak sections (re-enter Phase 4), maximum 2 cycles |
      | 0-59 | Major issues | Full rewrite (re-enter Phase 3), alert human reviewer |
      
      ## Auto-Revision Strategy
      
      When score is 60-69, target the weakest category first:
      
      1. **Low Content Quality:** Add missing topics from brief, inject specific examples into thin sections
      2. **Low SEO:** Fix keyword placement, add missing internal links, generate meta tags
      3. **Low E-E-A-T:** Add first-person experience language, inject specific data points, add source citations
      4. **Low Anti-Slop:** Run additional editing passes against banned phrase list, increase sentence length variation
      5. **Low GEO:** Restructure key sections with answer-first formatting, add FAQ schema, add source attributions
      
      **Maximum 2 revision cycles.** Research (Self-Refine) shows diminishing returns after 2 cycles, and quality can actually degrade with over-revision.
      
      ## Reporting Format
      
      ```markdown
      ## Quality Report
      
      **Overall Score: XX/100** [Ready | Minor Issues | Needs Revision | Major Issues]
      
      | Category | Score | Status |
      |----------|-------|--------|
      | Content Quality | XX/30 | [pass/flag] |
      | SEO Optimization | XX/25 | [pass/flag] |
      | E-E-A-T Signals | XX/15 | [pass/flag] |
      | Anti-Slop | XX/15 | [pass/flag] |
      | AI Citation Ready | XX/15 | [pass/flag] |
      
      **Readability:** Flesch XX (Grade Level X) -- [appropriate/too complex/too simple for target audience]
      
      **Issues Found:**
      1. [Specific issue + location in content + suggested fix]
      2. [...]
      
      **Strengths:**
      1. [What scored well]
      2. [...]
      ```
      
      ## Key Sources
      
      - [AgriciDaniel/claude-blog](https://github.com/AgriciDaniel/claude-blog) -- 100-point system across 5 dimensions
      - [aaron-he-zhu/seo-geo-claude-skills](https://github.com/aaron-he-zhu/seo-geo-claude-skills) -- 80-item CORE-EEAT framework
      - [hardikpandya/stop-slop](https://github.com/hardikpandya/stop-slop) -- 50-point scoring across 5 dimensions
      - [AI Content Quality Control Guide 2026 (Koanthic)](https://koanthic.com/en/ai-content-quality-control-complete-guide-for-2026-2/)
      - [Self-Refine: Iterative Refinement (Learn Prompting)](https://learnprompting.org/docs/advanced/self_criticism/self_refine)
      
    • voice-injection-playbook.md 16.5 KB
      # Technique 03: Voice & Personality Injection
      
      ## What It Is
      
      A system for injecting genuine human voice, personality, and specificity into AI-generated content. Removing AI patterns (Technique 02) is necessary but insufficient -- "sterile, voiceless writing is just as obvious as slop" (blader/humanizer). This technique adds the human qualities that make content engaging, trustworthy, and undetectable.
      
      ## Why It Works
      
      AI content is detectable primarily because it regresses to the mean of all writing. A specific voice has specific word preferences, sentence rhythms, opinions, and reference domains that break the low-perplexity pattern. Voice injection also directly addresses Google's E-E-A-T "Experience" signal -- content with genuine personality signals a real author with real opinions.
      
      Voice consistency is one of the strongest anti-AI-detection signals AND brand trust signals. Readers recognize when content "sounds different" -- it breaks trust. AI-generated content without voice injection sounds like every other AI-generated piece, which is detectable by both humans and algorithms.
      
      Google's EEAT framework implicitly rewards voice consistency: content from a recognizable, consistent voice signals a real author with real expertise. Generic AI voice signals generic AI.
      
      ## Step-by-Step Process
      
      ### Step 1: Voice Sample Collection
      1. Gather 3-5 real articles written by the brand's actual writer(s)
      2. Include variety: a formal piece, a casual piece, a technical piece
      3. Each sample should be 500-1,500 words minimum
      4. These samples are the SINGLE most important input for quality output
      5. **Capture negative examples too**: "We NEVER say 'leverage'" or "We NEVER use bullet points for everything" is just as important as positive guidelines
      
      ### Step 2: Voice Analysis
      6. Analyze the samples for:
         - **Sentence length patterns**: Does the writer use short, punchy sentences? Long, flowing ones? A mix?
         - **Vocabulary level**: Technical jargon? Plain language? Industry-specific terms?
         - **Tone markers**: Humorous? Direct? Academic? Conversational? Provocative?
         - **Structural preferences**: Lists vs prose? Subheadings frequency? Paragraph length?
         - **Personal pronouns**: "We" (corporate)? "I" (personal)? "You" (direct address)?
         - **Opening style**: Start with a story? A statistic? A question? A bold statement?
         - **Transition style**: Formal ("Furthermore")? Conversational ("But here's the thing")?
         - **Quirks**: Any distinctive patterns? Parenthetical asides? Rhetorical questions? Specific metaphors?
      
      ### Step 3: Voice Profile Creation
      7. Create a voice profile document with this structure:
      
      ```
      VOICE PROFILE: [Brand Name]
      
      Tone: [e.g., "Direct and slightly irreverent. Not afraid to call out bad advice."]
      Register: [e.g., "Professional but conversational -- like explaining to a smart colleague"]
      Perspective: [e.g., "First person plural ('we') for company content, first person singular ('I') for author-attributed pieces"]
      
      DO: [e.g., "Use short paragraphs, ask rhetorical questions, include specific numbers"]
      DON'T: [e.g., "Don't use corporate jargon, don't hedge everything, don't use emojis"]
      
      Vocabulary preferences:
      - Say "customers" not "clients"
      - Say "build" not "develop"
      - Say "test" not "validate"
      
      Example phrases that capture the voice:
      - "Here's the thing nobody talks about..."
      - "We tested this. The results surprised us."
      - "Skip the theory. Here's what actually works."
      ```
      
      ### Step 4: Voice Integration in Prompts
      8. Include the voice profile in every writing prompt
      9. Include 1-2 voice sample excerpts as few-shot examples
      10. Instruct the model: "Write as if you ARE [writer name]. Match their sentence patterns, vocabulary, and tone exactly."
      11. After generation, compare output against samples for voice drift
      
      ### Step 5: Voice Consistency Audit
      12. Read the draft next to a real sample -- do they sound like the same person?
      13. Check for AI voice bleeding through (formality spikes, generic transitions)
      14. Verify vocabulary consistency (using the brand's specific terms)
      15. Check opening and closing style matches the brand pattern
      16. **Update samples regularly**: If the brand's voice evolves (it will), update the samples. Stale samples produce dated-sounding content.
      
      ## The 10 Voice Injection Techniques
      
      ### 1. Named Persona
      
      **What:** Give the AI a specific writer identity with documented quirks, opinions, and speech patterns.
      
      **Mechanism:** Forces the model away from the statistical mean. A specific voice has specific word preferences, sentence rhythms, and opinions that break predictability.
      
      **Implementation:**
      ```
      You are writing as [Name], a [role] who has been doing this for [X years].
      They are known for: [specific trait, e.g., "blunt honesty about what doesn't work"]
      They frequently reference: [domain, e.g., "cooking analogies to explain technical concepts"]
      They dislike: [specific pet peeve, e.g., "vague advice without data"]
      They would never say: [banned phrases specific to this persona]
      Example of their voice: "[2-3 example sentences]"
      ```
      
      **Not:** "Write in an engaging, conversational tone." This is too vague and produces generic "friendly" AI writing.
      
      ### 2. Burstiness Enforcement
      
      **What:** Explicitly vary sentence length dramatically.
      
      **Mechanism:** Directly addresses the low-burstiness detection signal. Humans naturally produce "bursts" of complexity followed by simple punchy lines.
      
      **Implementation:**
      ```
      Vary sentence length dramatically. Follow a long, complex sentence with
      a short punchy one. Some paragraphs should be a single sentence. Others
      should be five or six sentences. Never let three consecutive sentences
      be similar in length. Mix 5-word fragments with 35-word complex sentences.
      ```
      
      **Example transformation:**
      - AI: "Email marketing remains one of the most effective digital marketing strategies available to businesses today. It provides a direct line of communication with potential customers. The return on investment is consistently high across industries."
      - Human: "Email marketing works. Not in the vague, 'it's part of a balanced strategy' way consultants love. I mean: for every euro you spend, you get 36 back. That's DKG's 2025 data across 4,000 campaigns, and it's been consistent for five years running."
      
      ### 3. Specificity Over Generality
      
      **What:** Replace every general claim with a specific example, number, name, date, or anecdote.
      
      **Mechanism:** Specific details are high-entropy tokens that break predictability. They also signal E-E-A-T (Experience, Expertise) to Google.
      
      **Implementation:**
      ```
      Never write a general claim without a specific example. Instead of
      "many companies struggle with this," write "When Coolblue's marketing
      team hit this wall in Q3 2025, they..." Every abstraction must be
      grounded in a concrete particular. If you don't have a real example,
      make the specificity about the scenario: exact numbers, exact steps,
      exact consequences.
      ```
      
      **The "30% Rule":** At least 30% of every page should contain details no generic AI could produce: proprietary data, original analysis, first-hand testing results, specific case studies, or named examples.
      
      ### 4. Contrarian/Opinionated Writing
      
      **What:** Take clear stances. Disagree with conventional wisdom when appropriate.
      
      **Mechanism:** AI is trained via RLHF to be balanced and inoffensive. Strong opinions are high-entropy by definition. They also drive engagement (comments, shares, backlinks) which are ranking signals via NavBoost.
      
      **Implementation:**
      ```
      Take a strong position on every topic. Don't hedge with "it depends"
      unless you genuinely mean it. If something is bad, say it's bad. If
      the conventional wisdom is wrong, say so and explain why with evidence.
      You are allowed to be wrong -- that's what makes you interesting.
      
      When appropriate, include phrases like:
      - "Here's what most guides get wrong about this..."
      - "I've tested this and the data says the opposite..."
      - "Everyone recommends X but in my experience Y works better because..."
      ```
      
      ### 5. Cultural References & Temporal Anchoring
      
      **What:** Reference specific cultural moments, current events, or shared experiences.
      
      **Mechanism:** Real cultural references signal an author who lives in the world. They create "information gain" that generic content cannot replicate. They anchor content in a specific time and place.
      
      **Implementation:**
      ```
      Reference specific cultural touchstones where relevant -- recent news,
      industry events, widely-known examples. Not as decoration but as genuine
      analogies that illuminate your point. Use references your target audience
      would know. For Dutch audiences: reference Prinsjesdag, Black Friday NL,
      specific Dutch companies, known media personalities, etc.
      ```
      
      ### 6. Show Don't Tell
      
      **What:** Demonstrate claims through mini-narratives and scenarios instead of stating them.
      
      **Mechanism:** "Telling" produces low-perplexity text ("Page speed affects rankings"). "Showing" produces high-perplexity, engaging text ("You click a search result. Three seconds pass. Still loading. You hit back. Google noticed.").
      
      **Implementation:**
      ```
      Never just state a fact. Show it through a brief scenario. Instead of
      "page speed affects rankings," write a mini-narrative: "You click a
      search result. Three seconds pass. The page is still loading. You hit
      back and click the next result. Google tracked every millisecond of that
      interaction -- and it just cost the slow site a ranking position."
      ```
      
      ### 7. Read-Aloud Test (Paul Graham Method)
      
      **What:** Every sentence should sound natural when spoken aloud to a friend.
      
      **Mechanism:** Paul Graham claims this puts you "ahead of 95% of writers." Conversational language has naturally higher burstiness, uses contractions, and avoids the formal register AI defaults to.
      
      **Implementation:**
      ```
      Write as if talking to a smart friend over coffee. Use contractions
      (don't, isn't, won't). Start sentences with "And" or "But" when natural.
      Drop unnecessary qualifiers. If you wouldn't say it out loud, rewrite it.
      
      Test: read every paragraph back. If it sounds like a textbook, rewrite
      it until it sounds like a person.
      ```
      
      ### 8. Strategic Imperfection
      
      **What:** Include deliberate human markers -- fragments, parenthetical asides, self-corrections.
      
      **Mechanism:** AI produces unnaturally perfect grammar and structure. Human writing has quirks that increase both burstiness and perplexity.
      
      **Implementation:**
      ```
      You are allowed to be imperfect. Use sentence fragments for emphasis.
      Start sentences with conjunctions. Use parenthetical asides (like this)
      when you have a tangential thought. Occasionally correct yourself
      mid-paragraph ("well, actually..." or "that's not quite right --").
      Real writers do this naturally.
      ```
      
      ### 9. Register Shifts
      
      **What:** Shift between formal and casual register within the same piece.
      
      **Mechanism:** AI maintains a uniform register. Humans naturally shift -- a technical explanation followed by a casual aside, a formal point followed by a joke.
      
      **Implementation:**
      ```
      Shift between registers naturally. After a technical paragraph,
      drop in something casual. After making a serious point, add a human
      aside. The contrast is what makes writing feel alive.
      
      Example: "The algorithm processes 12 trillion signals per query.
      (Yes, trillion. With a T. And it does this in under 0.5 seconds.
      Makes you feel productive, doesn't it?)"
      ```
      
      ### 10. The Gary Halbert "One Person" Principle
      
      **What:** Write to one specific person, not an audience.
      
      **Mechanism:** "Write your copy as if you're having a conversation with one reader." Using "I" and "you" creates intimacy. Specific, seemingly irrelevant details build trust.
      
      **Implementation:**
      ```
      Write to one specific person who already cares about this topic.
      Use "you" and "I" (or "we" for brand content). Include specific,
      concrete details even when they seem tangential -- they build trust.
      Don't persuade; serve. Imagine the reader is sitting across from
      you at a table, asking a real question.
      ```
      
      ## Voice Document Structure
      
      The writing skill should load a Voice Document (from your saved business context or user-provided) with this structure:
      
      ```markdown
      ## Brand Voice Profile
      
      ### Identity
      - Writer role/title:
      - Years of experience:
      - Known for:
      - Industry perspective:
      
      ### Voice Characteristics
      - Formality level: [casual / conversational / professional / formal]
      - Uses contractions: [yes / no]
      - Uses humor: [yes / sparingly / no]
      - Takes strong positions: [yes / sometimes / rarely]
      - Sentence length tendency: [short and punchy / mixed / flowing]
      
      ### Reference Domains
      - Analogies drawn from: [e.g., cooking, sports, construction]
      - Cultural references: [e.g., Dutch media, tech industry, specific era]
      
      ### Pet Peeves (things this voice NEVER does)
      - [e.g., "never says 'at the end of the day'"]
      - [e.g., "never hedges when they have data"]
      - [e.g., "never uses business jargon"]
      
      ### Example Paragraphs (2-3 samples of the target voice)
      [Actual writing samples that demonstrate the voice]
      
      ### Audience
      - Writing for: [specific person description]
      - Their knowledge level: [beginner / intermediate / expert]
      - What they care about: [specific concerns]
      ```
      
      **Voice varies by channel**: A brand's blog voice differs from their product page voice differs from their social media voice. Capture voice per content type if possible. The voice document should indicate which register applies to which content format.
      
      ## Copywriting Frameworks by Content Type
      
      | Content Type | Best Framework | Why |
      |-------------|---------------|-----|
      | Service pages, landing pages | PAS (Problem-Agitate-Solve) | Forces emotional engagement |
      | Commercial content, reviews | AIDA (Attention-Interest-Desire-Action) | Drives decision-making |
      | Brand content, about pages | StoryBrand | Positions reader as hero |
      | News, definitions, how-tos | Inverted Pyramid | Answer first, expand second |
      | Email-style, newsletters | Gary Halbert "One Person" | Maximum intimacy |
      | Data pages, comparisons | David Ogilvy "Facts First" | Specifics over superlatives |
      
      ## Tips
      
      - **The "swap test"**: Put a paragraph from the AI draft next to a paragraph from a real sample. If a colleague could tell which is which, the voice is not matched yet.
      - **Capture negative examples too**: "We NEVER say 'leverage'" or "We NEVER use bullet points for everything" is just as important as positive voice guidelines. What the brand avoids defines the voice as much as what it does.
      - **Voice varies by channel**: A brand's blog voice differs from their product page voice differs from their social media voice. Capture voice per content type.
      - **Update samples regularly**: If the brand's voice evolves (it will), update the samples. Stale samples produce dated-sounding content.
      - **The "30% Rule"**: At least 30% of every page should contain details no generic AI could produce.
      - **Voice before anti-slop**: Apply voice injection BEFORE running anti-slop audits (Technique 02). Removing patterns without adding personality produces sterile text.
      
      ## Common Mistakes
      
      1. **"Write engagingly"** -- too vague, produces generic friendly AI. Specify exact voice characteristics.
      2. **Using generic voice descriptions**: "Professional and friendly" describes 90% of brands. Be specific: "Slightly sarcastic, data-obsessed, uses 'actually' a lot"
      3. **Adding personality to every sentence** -- exhausting; let some sentences be plain
      4. **Forcing humor** -- bad jokes are worse than no jokes
      5. **Overusing fragments** -- occasional fragments add punch; constant fragments are a new AI pattern
      6. **Fake specificity** -- inventing plausible-sounding but false examples is worse than being general
      7. **Voice inconsistency** -- voice should be consistent throughout; load voice document before EVERY section
      8. **Over-constraining**: Too many rules make writing stiff. Focus on 5-7 key voice markers, not 50
      9. **Ignoring audience awareness**: The writer's voice adjusts to audience. A piece for CTOs sounds different than a piece for marketers, even in the same brand voice
      10. **One voice for all content**: Product descriptions, thought leadership, and technical docs require different registers within the same brand voice
      
      ## Key Sources
      
      - [Paul Graham: Write Like You Talk](https://paulgraham.com/talk.html)
      - [Gary Halbert: The Boron Letters (Enchanting Marketing)](https://www.enchantingmarketing.com/gary-halbert-boron-letters/)
      - [David Ogilvy's 7 Principles (Cult Method)](https://cultmethod.com/articles/ogilvys-principles/)
      - [StoryBrand Framework](https://storybrand.com/)
      - [blader/humanizer](https://github.com/blader/humanizer) -- "Sterile, voiceless writing is just as obvious as slop"
      - [haowjy/creative-writing-skills](https://github.com/haowjy/creative-writing-skills) -- Style learning from samples
      - [viktorbezdek/definitive-llm-writing-style-guide](https://github.com/viktorbezdek/definitive-llm-writing-style-guide)
      
    • writing-pipeline.md 15 KB
      # Technique 15: Writing Pipeline Architecture
      
      ## What It Is
      
      A multi-phase, multi-agent pipeline for producing high-quality SEO content. Based on analysis of 15+ agentic writing systems, this architecture represents the consensus best approach: specialized phases with quality gates, not a single monolithic generation.
      
      ## Why It Works
      
      Research consistently shows that multi-agent pipelines outperform single-prompt generation:
      - Section-by-section generation maintains quality over long articles
      - Specialized agents (writer vs. critic) produce better results than a generalist
      - One to two revision cycles capture 95% of improvement (Self-Refine research)
      - Quality gates prevent bad content from being output
      
      Human-written articles still generate 5.44x more traffic and hold attention 41% longer than pure AI. The pipeline must close this gap through specialization.
      
      Single-prompt content generation fails because one prompt can't simultaneously optimize for research depth, writing quality, SEO mechanics, anti-detection, and fact accuracy. Each agent focuses on ONE thing well, and each stage builds on verified output from the previous stage.
      
      ## The 7-Agent Model
      
      The pipeline uses seven specialized agents, each with a distinct role and optimal configuration:
      
      | Agent | Role | Temperature | Tools |
      |-------|------|-------------|---------------|
      | Researcher | Analyze topic landscape, SERP competition, information gaps | Low (0.2-0.3) | keyword search, competitor data, SERP feature detection |
      | Outliner | Create SERP-optimized content structure | Low-Medium (0.3-0.4) | opportunity detection, keyword cluster data |
      | Writer | Generate first draft with anti-detection built in | Medium (0.5-0.7) | page content rendering (competitor analysis) |
      | Humanizer | Audit and fix surviving AI patterns | Low (0.2-0.3) | None |
      | Fact-Checker | Verify all claims, statistics, attributions | Low (0.1-0.2) | None |
      | SEO Optimizer | Keyword placement, meta elements, schema, linking | Low (0.2-0.3) | internal link data, SERP feature detection |
      | Quality Reviewer | Final holistic review and scoring | Low (0.2-0.3) | None |
      
      ## Implementation Architecture
      
      ```
      +--------------+     +----------+     +--------+     +-----------+
      | RESEARCHER   | --> | OUTLINER | --> | WRITER | --> | HUMANIZER |
      +--------------+     +----------+     +--------+     +-----------+
                                                                |
      +--------------+     +----------+     +----------------+  |
      | QUALITY      | <-- | SEO      | <-- | FACT-CHECKER   | <+
      | REVIEWER     |     | OPTIMIZER|     |                |
      +--------------+     +----------+     +----------------+
            |
            v
        [FINAL OUTPUT]
        - Article (markdown)
        - Meta elements
        - Quality scorecard
        - SEO checklist
        - Schema JSON-LD
        - AI-detection risk score
      ```
      
      ## User Approval Workflow
      
      The pipeline is NOT fully autonomous. Human checkpoints are mandatory:
      
      1. User invokes "write content" -> triggers RESEARCHER
      2. RESEARCHER output presented to user for review/approval
      3. OUTLINER generates structure -> user approves or modifies
      4. Pipeline PAUSES to collect human input (experiences, data, opinions)
      5. WRITER generates with all context -> HUMANIZER audits
      6. FACT-CHECKER flags issues -> user verifies first-party claims
      7. SEO OPTIMIZER + QUALITY REVIEWER finalize
      8. Final output with scorecard presented to user
      
      At every human checkpoint, the user can modify, reject, or redirect. This is not optional overhead -- it is the mechanism that prevents high-quality slop.
      
      ## Pipeline Architecture (Detailed Phases)
      
      ### Phase 1: Preparation (Context Loading)
      
      **Inputs loaded:**
      1. Content Brief (from Content Brief Generator)
      2. Voice Document (from your saved business context or user-provided)
      3. Content Type Template (selected based on intent/type from brief)
      4. Anti-Slop Rules (banned phrases, structural patterns)
      5. Internal Link Map (from internal link data)
      6. Competitor Content Analysis (from page content rendering on top competitor URLs)
      
      **Critical:** All context must be loaded BEFORE writing begins. Loading mid-article breaks voice consistency.
      
      **The RESEARCHER agent performs:**
      1. Queries tools: keyword search, competitor data, SERP feature detection
      2. Analyzes top 10 SERP results for the target keyword
      3. Maps competitor content coverage (topics covered, depth, unique angles)
      4. Identifies information gaps (topics nobody covers)
      5. Notes SERP features present (PAA, featured snippets, video carousels)
      6. Recommends target word count based on competitor analysis
      
      **Output:** Research brief with SERP gap analysis, competitor coverage map, information gain opportunities.
      
      ### Phase 2: Outline Generation
      
      **The OUTLINER agent generates a detailed outline from the content brief:**
      1. SERP-derived H2/H3 structure (matching what ranks)
      2. Keywords mapped to specific sections
      3. Internal links planned for specific paragraphs
      4. Data points / examples planned per section
      5. Featured snippet target identified (format + section)
      6. PAA questions assigned to sections
      7. Word count allocation per section
      8. "Human input required" points marked (case studies, data, opinions)
      
      **Human approval gate:** Present outline for approval before proceeding. In automated mode, validate against brief structure. The user can add, remove, or reorder sections.
      
      ### Phase 3: Section-by-Section Writing
      
      **Why section-by-section:** For articles over 1,500 words, full-article generation leads to quality drift, repetition, and context distraction. Section-by-section maintains focused context.
      
      **The WRITER agent processes each section:**
      1. Load section brief (H2/H3 structure, assigned keywords, planned links, planned examples)
      2. Load voice document (reload for each section to maintain consistency)
      3. Apply content type template rules
      4. Inject human-provided content at designated points
      5. Write the section with anti-detection built in (varied sentence length 3-40 words, Tier 1 word avoidance)
      6. Place internal links at natural anchor points
      7. Update scratchpad with key points covered (prevents repetition)
      8. Compress section summary for context management
      
      **Context management:** After each section, create a summary of what has been covered. Use the summary (not full text) as context for subsequent sections. This prevents the model from losing focus as context grows.
      
      **Key finding from Anthropic:** Model performance drops around 32,000 tokens even with million-token windows due to context distraction. Use "just-in-time context" -- lightweight identifiers that dynamically load data at runtime, not monolithic context blocks.
      
      **Section compression strategy:** Each completed section is reduced to a 2-3 sentence summary containing: key points made, data cited, keywords used, links placed. The full text is stored separately. The writer for the next section sees only these summaries plus the current section's brief.
      
      ### Phase 4: Anti-Slop Editing
      
      **Separate from writing.** The writer agent should be "naive" about quality rules (DonAldente pattern) -- its job is to write naturally. The editor/critic is separate and ruthless.
      
      **The HUMANIZER agent runs two passes:**
      
      **Pass 1: Detection**
      - Scan for Tier 1 banned phrases (always remove)
      - Scan for Tier 2 clustering (flag if 3+)
      - Scan for Tier 3 density (flag clusters)
      - Check structural patterns (uniform sentence length, rule-of-three, etc.)
      - Check burstiness (sentence length standard deviation must be > 5 words per paragraph)
      - Check rhythm uniformity (flags sections where all sentences are 15-25 words)
      - Score total points
      
      **Rewrite:** For each flagged section, rewrite preserving meaning while eliminating the pattern. Do not just delete -- replace with human-sounding alternatives.
      
      **Pass 2: Surviving Pattern Check**
      - Re-scan the full text after Pass 1 rewrites
      - First-pass rewrites retain 15-20% of AI patterns
      - Specifically check for recycled transitions, lingering inflation, new uniformity
      
      **Horoscope Test:** For each paragraph, ask: "Could anyone have written this, for anyone?" If yes, flag for specificity injection.
      
      **Critical threshold:** If the Humanizer has to rewrite more than 20% of the content, the Writer prompt needs improvement. The Writer should produce mostly clean output; the Humanizer is a quality gate, not the primary fix.
      
      ### Phase 5: Fact-Checking
      
      **The FACT-CHECKER agent verifies:**
      1. All factual claims in the content
      2. Statistics against named sources
      3. Fabricated or unverifiable claims (flag for removal)
      4. All "experts say" have named experts
      5. URLs and references are real and current
      6. Claims that need human verification (first-party data)
      
      **Output:** Fact-checked draft + verification report with confidence levels per claim.
      
      ### Phase 6: SEO Optimization
      
      **The SEO OPTIMIZER agent handles:**
      
      **Keyword checks:**
      - Primary keyword in meta title, H1, first paragraph, 2-3 H2s
      - ~2% body density (naturally distributed, not forced)
      - Keyword variations in H3s and image alt text
      
      **Internal link validation:**
      - Every internal link must point to a real, existing page
      - Validate against internal link data
      - Use descriptive anchor text (not "click here")
      - 3-5 contextual links per 1,000 words
      
      **Meta generation:**
      - Meta title: primary keyword + benefit/hook, under 60 characters
      - Meta description: value proposition + CTA, under 155 characters
      - 5 title variations, 3 description variations
      
      **Schema markup:**
      - Generate JSON-LD based on content type (HowTo, FAQ, Product, etc.)
      - Include FAQPage schema for all content with FAQ sections
      - Validate schema structure
      
      **Featured snippet targeting:**
      - Verify the target snippet section has a 40-60 word direct answer
      - Format matches expected snippet type (paragraph/list/table)
      
      ### Phase 7: Quality Scoring
      
      **The QUALITY REVIEWER agent performs a holistic review and scores on a 100-point composite:**
      
      | Category | Points | Components |
      |----------|--------|------------|
      | Content Quality | 30 | Topical completeness (10), Depth/examples per section (10), Originality (10) |
      | SEO Optimization | 25 | Keyword placement (8), Internal links verified (5), Meta quality (4), Schema (4), Snippet targeting (4) |
      | E-E-A-T Signals | 15 | First-person experience (5), Specific data/examples (5), Source attribution (5) |
      | Anti-Slop Score | 15 | Zero Tier 1 phrases (5), Burstiness (3), Voice consistency (3), Horoscope Test (4) |
      | AI Citation Ready | 15 | Answer-first formatting (5), FAQ schema (3), Claim-evidence pairs (4), Entity richness (3) |
      
      **The reviewer also checks:**
      - Flow and logical progression across the full article
      - Voice consistency throughout (not just per-section)
      - Information gain elements are present and effective
      - EEAT signals are woven in (not bolted on)
      
      **Quality gates:**
      - Score < 60: Full rewrite (re-enter Phase 3)
      - Score 60-69: Auto-revise weak sections (re-enter Phase 4 for flagged sections)
      - Score 70-84: Flag issues for human review, output with annotations
      - Score 85+: Ready for publication
      
      **Auto-revision:** Target the weakest scoring category first. If Anti-Slop scored low, run additional editing passes. If E-E-A-T scored low, inject experience signals. Maximum 2 revision cycles (research shows quality degrades beyond this).
      
      ## Key Design Decisions
      
      ### Sequential, Not Parallel
      Each agent needs the previous agent's output. The Writer can't write without the Outline. The Humanizer can't audit without the draft. Parallel execution produces worse results because agents lack context. CrewAI and LangGraph implementations have confirmed that sequential pipelines produce higher quality.
      
      ### Context Passing Is Full, Not Summarized
      Each agent receives the FULL output of the previous agent, not a summary. Summaries lose critical nuances (specific data points, exact quotes, structural decisions). The exception is section-to-section context within the Writer phase, where compression prevents context distraction.
      
      ### Anti-Detection Is Built In, Not Bolted On
      The Writer already follows anti-detection guidelines. The Humanizer is a QUALITY GATE, not the primary fix. If the Humanizer has to rewrite more than 20% of the content, the Writer prompt needs improvement.
      
      ### Human Input Is Required, Not Optional
      The pipeline HALTS at the Writer stage to request human input (case studies, data, opinions). Without this input, the pipeline produces content with zero information gain -- which defeats the entire purpose.
      
      ## Output Format
      
      ```markdown
      ---
      title: [Generated meta title]
      description: [Generated meta description]
      keywords: [Primary keyword, secondary keywords]
      schema_type: [HowTo | FAQ | Article | Product | etc.]
      word_count: [Actual word count]
      quality_score: [0-100]
      tarsnippet data: [paragraph | list | table | none]
      ---
      
      # [H1 with primary keyword]
      
      [Article content in Markdown]
      
      ---
      <!-- Internal links used: [list of URLs] -->
      <!-- Schema JSON-LD: [embedded or separate file] -->
      <!-- Quality breakdown: Content X/30, SEO X/25, E-E-A-T X/15, Anti-Slop X/15, GEO X/15 -->
      ```
      
      ## Systems Studied
      
      | System | Key Innovation | Adopted |
      |--------|---------------|---------|
      | SEO Machine | Editor agent targeting AI patterns | Yes -- Phase 4 |
      | DonAldente anti-slop | Naive drafter + ruthless critic | Yes -- Phase 3 vs 4 separation |
      | claude-blog | Dual Google + AI citation scoring | Yes -- Phase 7 scoring |
      | Agentic-SEO-Skill | Confidence labeling on claims | Adapted -- source attribution |
      | AutoGen Reflection | JSON-summarized multi-reviewer | Adapted -- structured quality scoring |
      | Self-Refine research | 1-2 revision cycles optimal | Yes -- max 2 revision cycles |
      | Anthropic context engineering | Just-in-time context loading | Yes -- section compression |
      
      ## Common Mistakes
      
      1. **Running all agents with the same LLM temperature**: Researcher needs low temperature (factual), Writer needs medium (creative), Humanizer needs low (precise editing). See the agent table above for recommended values.
      2. **Skipping the Fact-Checker**: AI confidently fabricates statistics. One fake stat destroys the entire article's credibility.
      3. **Making the pipeline fully autonomous**: Without human input, the pipeline produces high-quality AI slop. Human input at the Writer stage is the critical differentiator.
      4. **Post-processing instead of pre-processing**: Anti-detection works 5x better when built into the writing prompt vs applied after.
      5. **Monolithic context loading**: Loading all context at once causes performance degradation past 32K tokens. Use section compression and just-in-time context instead.
      6. **Over-revising**: More than 2 revision cycles produces diminishing returns and can actually degrade quality. Fix the upstream prompt instead.
      
      ## Key Sources
      
      - [TheCraigHewitt/seomachine](https://github.com/TheCraigHewitt/seomachine)
      - [DonAldente-AI/anti-slop-system](https://github.com/DonAldente-AI/anti-slop-system)
      - [AgriciDaniel/claude-blog](https://github.com/AgriciDaniel/claude-blog)
      - [Self-Refine: Iterative Refinement (Learn Prompting)](https://learnprompting.org/docs/advanced/self_criticism/self_refine)
      - [Effective Context Engineering (Anthropic)](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents)
      - [Building Effective Agents (Anthropic)](https://www.anthropic.com/research/building-effective-agents)
      
  • SKILL.md 5.9 KB
    ---
    name: improve-content
    description: Use when rewriting or refreshing an existing page that's underperforming. The agent fetches the URL, analyzes the current content, researches the SERP, and rewrites using the full anti-AI-slop ruleset — no data exports needed.
    ---
    
    # Improve Content
    
    Rewrites an existing page so it outranks competitors — better structure, deeper expertise, and a voice that actually sounds human. Uses the same anti-slop ruleset as `write-content`.
    
    ## Input
    
    **URL of the page to improve**. If the fetch fails, ask the user to paste the page content directly.
    
    ## Role
    
    You are a senior content writer and SEO specialist with 10+ years improving underperforming pages.
    
    ## Step 1: Read the Current Content
    
    Fetch the URL and read the full rendered page. Note title, meta, H1/H2s, word count, internal links, schema. Identify the apparent primary keyword from the content.
    
    ## Step 2: Research the SERP
    
    Google the primary keyword. Read the top 5 results. Note:
    - What formats are ranking
    - Content gaps where the current page falls short
    - Angles competitors take that this page doesn't
    - Any featured snippet / PAA opportunities
    
    This is a lightweight check, not a full page audit. Just enough to know what's missing.
    
    ## Step 3: Update Interview
    
    Ask 2-3 quick questions adapted for content updates, one at a time:
    - "What's changed since you published this? New data, new experience, market shifts?"
    - "What results did this content get? Traffic, leads, feedback?"
    - "Knowing what you know now, what would you add or cut?"
    
    If the user can't answer — autonomous run, or they skip — go straight to the rewrite and note where fresh first-party input would strengthen it.
    
    ## Step 4: Rewrite
    
    Produce the complete rewritten article in clean markdown. Apply ALL writing rules:
    
    ### Voice
    Practitioner to a peer. Take positions. Use "you" and "I/we." Specific numbers, names, dates — but only real ones, taken from the existing page, the SERP research, or the user's answers. Never invent a number, name, or citation; cut the claim instead. Contractions. Show thinking changing. Shift registers — after technical explanation, drop into casual aside. Uniform register = AI tell.
    
    ### Rhythm
    Vary sentence length dramatically (5 to 30+ words). Vary paragraph length. Fragments. Parenthetical asides. Break the topic-sentence-support pattern. Cover sections asymmetrically. No section summaries.
    
    ### Show, don't state
    Brief scenarios instead of flat claims. "You click a search result. Three seconds pass. Still loading. You hit back." narrates; "page speed affects UX" states.
    
    ### Anti-slop
    
    **NEVER use**: delve, landscape (metaphorical), testament, leverage, utilize, robust, seamless, furthermore, moreover, additionally, pivotal, multifaceted, harness, embark, navigate (metaphorical), showcase, streamline, paramount, culminate, spearhead, commence, endeavor, vibrant, innovative, comprehensive (as adjective).
    
    **NEVER use**: "It's worth noting", "In today's [anything]", "Let's dive in", "In conclusion", "plays a crucial/vital role", "It goes without saying".
    
    **Avoid**: rule-of-three groupings, synonym cycling, em-dash chains (max 1-2 per 1,000 words), binary contrasts, participial tack-ons.
    
    ### SEO
    Primary keyword in H1, first 100 words, 2-3 H2s. ~2% density, naturally distributed. 40-60 word direct answer after the most important H2 (featured snippet). Weave PAA questions as H2/H3 headings. 3-5 internal links per 1,000 words. Front-load value.
    
    ### Content type
    
    Detect from the existing page + SERP. Load `references/content-types-overview.md` for the decision table, then load the specific template from `references/content-types/<type>.md` for the full rewrite structure. 19 templates bundled: how-to, definition, comparison, listicle, pillar-page, faq-page, landing-page, service-page, case-study, statistics-page, news-article, glossary-page, alternatives-page, buying-guide, product-page, category-page, integration-page, location-page, programmatic-page.
    
    - **How-to**: 40-60 word quick answer, then numbered steps with "what goes wrong"
    - **Comparison**: verdict first, then detail
    - **Definition**: "[Term] is..." in the first sentence
    - **Case study**: result number first, then the story
    
    ### Final checks
    1. "So what?" test per section — could anyone have written this for anyone? If yes, inject specific knowledge.
    2. Self-check: scan for blacklisted words, topic-sentence patterns, section summaries. Fix before delivering.
    
    ### Output
    Clean markdown. Title + rewritten article content. Nothing else.
    
    ### Language
    Write in the same language as the existing page content.
    
    ## Bundled references
    
    Load from `references/` only when needed — don't preload the whole folder.
    
    **Content type templates** (`references/content-types/`) — load one in Step 4:
    - Common: `how-to.md`, `definition.md`, `comparison.md`, `listicle.md`, `pillar-page.md`, `faq-page.md`, `landing-page.md`, `service-page.md`, `case-study.md`
    - Content and news: `statistics-page.md`, `news-article.md`, `glossary-page.md`
    - Commercial: `alternatives-page.md`, `buying-guide.md`, `product-page.md`, `category-page.md`, `integration-page.md`, `location-page.md`, `programmatic-page.md`
    - `references/content-types-overview.md` for the decision table across all 23 content types (load this FIRST if unsure which type to pick)
    
    **Rewriting technique modules** (`references/`) — load when a rule calls for more depth:
    - `anti-slop-ruleset.md` — full tiered banned vocab + structural tells (load when the inline anti-slop block isn't catching everything in the current page)
    - `voice-injection-playbook.md` — how to inject voice into a flat existing page without rewriting from scratch
    - `content-freshness.md` — decay signals and what to refresh vs. leave alone (the core of Step 3's update interview)
    - `writing-pipeline.md` — the research → rewrite → edit loop adapted for updates
    - `quality-scoring.md` — scoring rubric to run before delivering the rewrite
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.

Related