agentleFS
narayan-metaflownarayan-metaflow/metaflow-marketing-skills/skills/seo-aeo-content-strategy/SKILL.md

seo-aeo-content-strategy

Skill14 starsChanged 7 months agoMITView on GitHub

Design content systems that earn both Google rankings *and* AI citations. The core insight: AI answer engines read content differently than humans — they extract the highest-information-gain sentences and treat H2 headings as prompts. The \"ski ramp\" structure optimizes for both.

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 12 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    All citations above are data-backed.

    line 17
  2. Very specific:

    Create content briefs with AEO structure baked in: Use the ski ramp framework by default.

    line 25
  3. Distinctive:

    Every cluster page links back to the pillar (upward)

    line 45
  4. Distinctive:

    Put the key claim, definitive answer, or headline finding in the first paragraph.

    line 70
  5. Distinctive:

    No throat-clearing intros.

    line 71
  6. Distinctive:

    Write the way a journalist writes a news lede: most important fact first.

    line 72
  7. Distinctive:

    No hedging. "Teams that publish twice a week see 2.3x more traffic" beats "Publishing frequency may help traffic."

    line 80
  8. Distinctive:

    Use specific brand names, tool names, numbers, dates, proper nouns.

    line 85
  9. Distinctive:

    Put the information gain — the specific, concrete, citable fact — somewhere in the middle of a 3-5 sentence paragraph, after context and …

    line 91
  10. Distinctive:

    Don't artificially stuff entities or brand names — modern LLMs detect this and it damages trust signals.

    line 98
  11. Distinctive:

    Don't build content just to be "summarizable." Large-sample work: making content summarizable alone doesn't make earning citations easier.

    line 99
  12. Distinctive:

    Don't use clickbait H2s that don't match the question a user would actually ask.

    line 100
  13. Distinctive:

    When planning content briefs or writing to the standard, read:

    line 267

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Why this matters

  • Topic clusters drive ~30% more organic traffic and hold rankings 2.5x longer than standalone content (industry benchmarks, 2025).
  • 44.2% of all LLM citations come from the first 30% of text (large-sample ChatGPT citation analysis, ~1.2M pages). The intro carries disproportionate weight — front-load your answer.
  • 78.4% of citations containing questions come from H2 headings — LLMs treat H2s as prompts and the following paragraphs as answers.
  • Heavily-cited text has ~20.6% entity density (brand names, tool names, proper nouns) vs. 5-8% for normal English.
  • 53% of citations come from the middle of a paragraph, not the first sentence. LLMs extract the sentence with the highest information gain, not the opener.
  • Content not updated quarterly is ~3x more likely to lose AI citations (multi-model freshness benchmarks).
  • Brands are 6.5x more likely to be cited via third-party sources than their own domains (vendor research, Oct 2025). Distribution matters as much as publication.
  • Earned media distribution increases AI citations up to 325% vs publishing only on your own site (syndication research, Dec 2025).

All citations above are data-backed. The ski ramp framework has p-value 0.0 (statistically indisputable) in large-sample citation studies.

Process

  1. Define business goals and audience. What must the content produce — leads, trials, awareness, authority? Who specifically is the reader, in which intent state?
  2. Map the topic universe. Core topics (3-7) → subtopics (5-15 per core) → supporting questions (unlimited). This becomes the cluster architecture.
  3. Design cluster architecture. Pillar pages (broad, 2500+ words, link hub) → cluster pages (specific subtopics, link back to pillar + peers) → supporting content (FAQ, glossary, tools, comparisons).
  4. Audit existing content. Score every page on the 1-5 matrix below: keep, update, consolidate, prune, or identify as a gap to create.
  5. Create content briefs with AEO structure baked in. Use the ski ramp framework by default. See references/content-brief-template.md for the complete brief template.
  6. Build the editorial calendar. Mix new creation with refresh cycles. Assume quarterly review of all priority content.
  7. Define measurement criteria. What does success look like per content type? Rankings, traffic, AI citations, leads, revenue?

Frameworks

Topic cluster design template

Pillar page: "The Complete Guide to [Core Topic]"
├── Cluster page 1: "[Specific subtopic]"
├── Cluster page 2: "[Specific subtopic]"
├── Cluster page 3: "[Specific subtopic]"
│   ├── Supporting: FAQ
│   ├── Supporting: Comparison / vs
│   └── Supporting: Tool/calculator
└── Cluster page N: ...

Linking rules:

  • Every cluster page links back to the pillar (upward)
  • Pillar links to all clusters (downward)
  • Clusters cross-link to 2-3 topically related clusters (sideways)
  • Supporting content links to its parent cluster
  • Anchor text is descriptive, not "click here"

Sizing: pillar pages 2,500-5,000 words; cluster pages 1,200-2,500 words; supporting content as needed. These are guides, not targets — write what the topic actually requires.

Content audit scoring matrix

ScoreLabelActionCriteria
1StrongKeep / MonitorRanking top 10, driving conversions, recently updated, AI-cited
2Needs refreshUpdateRanking 11-30, traffic declining, outdated info, structurally sound
3DuplicativeConsolidateMultiple thin pages on same topic, cannibalizing each other
4UnderperformingPrune or rewriteNot ranking, no traffic, no backlinks, thin content
5GapCreateTopic covered by competitors but missing from our site

Pruning is not deleting. A pruned page may be merged into another (301 redirect), rewritten entirely, noindexed (for legal/historical retention), or deleted with a redirect to the most relevant parent page.

The "Ski Ramp" — AEO content structure

This is the single most important content framework in this skill. Derived from large-sample ChatGPT citation analysis (p-value 0.0). Apply it to every piece of content meant to earn AI citations.

1. Front-load the answer (first 30% captures 44.2% of citations).

  • Put the key claim, definitive answer, or headline finding in the first paragraph.
  • No throat-clearing intros. No "In this article, we'll explore…"
  • Write the way a journalist writes a news lede: most important fact first.

2. Use question-based H2 headings (78.4% of question-containing citations come from headings).

  • LLMs treat H2s as prompts and the following paragraphs as answers.
  • Phrase H2s the way a user would ask the question: "How long does SEO take to show results?" not "SEO Timeline Overview."
  • One topic per H2 section.

3. Follow each H2 with a direct, definitive answer.

  • No hedging. "Teams that publish twice a week see 2.3x more traffic" beats "Publishing frequency may help traffic."
  • The direct answer goes in the first 1-2 sentences after the heading. Supporting detail follows.
  • LLMs penalize hedging language in the extraction pass.

4. High entity density (~20%).

  • Use specific brand names, tool names, numbers, dates, proper nouns.
  • "Increase budget by 15% in Q3" beats "increase budget in the second half of the year."
  • Dates and numbers are universal citation positives — include real ones whenever possible.

5. The middle of paragraphs matters.

  • 53% of citations come from mid-paragraph, not the opening sentence.
  • Put the information gain — the specific, concrete, citable fact — somewhere in the middle of a 3-5 sentence paragraph, after context and before the transition.

6. Include a summary/conclusion section (the "wake-up" zone).

  • The last 10% of content before the footer sees a citation bump.
  • A short summary or "Key takeaways" block captures this.

What NOT to do:

  • Don't artificially stuff entities or brand names — modern LLMs detect this and it damages trust signals.
  • Don't build content just to be "summarizable." Large-sample work: making content summarizable alone doesn't make earning citations easier. Information gain and entity clarity do.
  • Don't use clickbait H2s that don't match the question a user would actually ask. The AI matches on semantic intent, not clever copy.

Content brief template (summary)

Core fields every brief should include:

FieldNotes
Target keyword clusterPrimary + 5-15 supporting
Search intentInformational / commercial / transactional / navigational
Fan-out sub-queriesFrom seo-keyword-research skill
Target word countBased on top-10 SERP average + 20%
H1Primary topic, natural phrasing, includes target KW
H2 structureQuestion-based, one topic each, sequenced by the user journey
First-paragraph key claimThe specific answer/finding to front-load
Key entitiesNamed brands, tools, people, places, dates, numbers to include
Internal linksTo pillar (required) + 2-3 related clusters + supporting content
External citationsPrimary sources, data, expert quotes
Schema typeArticle, FAQPage, HowTo, Product, etc.
Competitor content to beatTop 3-5 currently ranking
Information gain angleWhat new data/insight do we add that nothing else has?
Named author + credentialsRequired for E-E-A-T

See references/content-brief-template.md for the complete version with prompts and guidance for each field.

Content freshness cadence

Content typeRefresh intervalTriggers an ad-hoc refresh
Pillar pagesQuarterlyCore update, major competitor launch, outdated stat
Cluster pages (priority)QuarterlyDrop below top 10, traffic decay >20%, factual staleness
Cluster pages (secondary)Semi-annuallySignificant SERP shift
News/trend postsAs neededOriginal news posts do not get refreshed; they get superseded
Evergreen referenceAnnuallyOnly when facts change
Product/pricing pagesOn changeWhenever pricing, features, or positioning shifts

Why: Content not updated quarterly is ~3x more likely to lose AI citations in published multi-model benchmarks. Freshness is a confirmed signal across several open and closed LLMs. This is a real, data-backed signal — but remember, AI citation itself is volatile (45.5% of citations change between consecutive observations), so absolute consistency is impossible.

E-E-A-T checklist

Applies to Google quality rater guidelines and overlaps with LLM trust signals.

  • Named author with real credentials (not "Admin" or "Editorial Team")
  • Author bio / about page with professional history
  • First-party expertise, original research, or direct experience cited
  • Sources linked to primary data (not other blogs citing the same source)
  • Publication date visible on-page
  • Last-updated date visible when content is refreshed
  • Clear editorial process or methodology statement
  • Organization schema with founding date, logo, social profiles
  • Author schema (Person type) linking to credentials

Information gain strategy

"Information gain" = what does this page add that no other page on the topic contains? Without it, you're competing on formatting and word count, which is a losing game against established sites.

High information gain sources:

  • Original research (surveys, studies, experiments)
  • First-party data (usage metrics, customer anonymized data, proprietary analytics)
  • Expert commentary from named practitioners
  • Proprietary methodology or framework
  • Case studies with real numbers
  • Specific screenshots, walkthroughs, or demonstrations nobody else has

Critical amplification insight: Brands are 6.5x more likely to be cited via third-party sources than their own domains. Earned media distribution boosts AI citations up to 325% vs publishing only on-site. This means original research has more value when distributed — a proprietary study pitched to industry publications, syndication partners, and expert podcasts will earn more AI citations than the same study published only on your blog.

Plan information gain as a distributable asset, not a blog post. Hand off to seo-link-building for the distribution plan.

Output format

## Content Strategy — [Client/Project]

**Business goals:** [List]
**Target audience:** [Persona + intent state]
**Time horizon:** [Quarter / half-year / year]

### Topic universe
| Core topic | Subtopics | Pillar status | Cluster status |
|-----------|-----------|---------------|----------------|

### Cluster architecture
[Visual tree or table showing pillar → cluster → supporting for each core topic]

### Content audit summary
| Score | Count | Action | Est. effort |
|-------|-------|--------|-------------|
| 1 Keep | | | |
| 2 Update | | | |
| 3 Consolidate | | | |
| 4 Prune/rewrite | | | |
| 5 Gap — create | | | |

### Editorial calendar (next 90 days)
| Week | Action | Content | Type | Brief owner | Writer |
|------|--------|---------|------|-------------|--------|

### Content briefs
[Link or reference to detailed briefs for each piece — see references/content-brief-template.md]

### Information gain / distribution plan
[What original assets will we create? How will they be distributed beyond our site?]

### Measurement
| KPI | Baseline | Target | Review cadence |
|-----|----------|--------|----------------|

Example — SaaS analytics startup

Goals: 3x organic leads in 6 months; establish topical authority in product analytics. Audience: Product managers and growth leads at B2B SaaS companies, 10-200 employees. Time horizon: Next 90 days (with 6-month vision).

Topic universe (core topics):

  1. Product analytics fundamentals
  2. Event tracking and instrumentation
  3. Funnel analysis
  4. Retention and cohort analysis
  5. A/B testing and experimentation

Cluster architecture for "Product analytics fundamentals" (example):

  • Pillar: /product-analytics-guide (3,500 words, the definitive primer)
    • Cluster: /product-analytics-vs-web-analytics
    • Cluster: /product-analytics-tools-comparison
    • Cluster: /how-to-choose-product-analytics-tool
    • Cluster: /product-analytics-metrics-that-matter
    • Supporting: /product-analytics-glossary

Audit (36 existing posts):

  • 8 Strong — keep as-is
  • 14 Needs refresh (mostly 2023 posts with outdated stats)
  • 6 Consolidate into 2 pillar pages
  • 5 Prune (thin, no traffic, no backlinks)
  • 12 Gaps identified against 3 competitor sites

Ski ramp brief for /product-analytics-guide pillar:

  • First paragraph key claim: "Product analytics tools capture user interactions with your product to answer three questions: what users do, why they do it, and what to change next. Teams that instrument 15-30 core events see 2-3x better retention visibility than teams tracking everything."
  • H2s (all question-based):
    • What is product analytics?
    • How is product analytics different from web analytics?
    • What events should I track?
    • Which product analytics tool is right for my team?
    • How do I measure retention with product analytics?
    • What does a product analytics stack look like in 2026?
  • Key entities: named tools in the category, GA4, event, funnel, cohort, retention, activation (adjust to your vertical).
  • Information gain angle: Original benchmark survey of 240 product teams on instrumentation depth vs retention visibility.
  • Distribution plan: Syndicate the benchmark to Mind the Product, ProductLed, and offer exclusive cut to one Tier-1 publication for launch day coverage.

Editorial calendar: 6 new pieces / month + 4 refreshes / month. Benchmark survey is the Q2 tentpole asset.

Guidelines

  • Apply the ski ramp to every page meant to earn AI citations. It is the highest-confidence AEO framework we have (large-sample citation analysis, p=0.0).
  • Never hedge in the first paragraph. "This may help teams understand X" ranks and cites worse than "Teams that do X see Y." Definitive language wins in both humans and LLMs.
  • Entity density is a knob, not a dial. Target ~20% entity density naturally — don't stuff. If it reads awkwardly to a human, it reads awkwardly to the ranking systems.
  • Pillar pages cannot be shallow. 2,500+ words is typical because a true pillar answers the full topic. If you can't write 2,500 words of substance, it's not a pillar topic for you yet.
  • Audit before you build. Most content budgets are wasted on net-new pieces when existing content would outperform with a refresh. Score before you write.
  • Information gain is leverage. Original research is 10x the work of rehashing; if distributed well, it's 100x the return. Budget for both creation and distribution.
  • Quarterly refresh is the default cadence, not a best-case. Build it into the calendar from day one.
  • E-E-A-T is not a checkbox. "Add an author photo" without real credentials is theater. The underlying trust signal has to be real — named expert, actual experience, verifiable history.
  • Do not believe the "AI content penalty" myth. Large-sample citation studies (~1.9M pages) — 87.8% of AI-cited pages are at least AI-assisted. Quality and information gain matter; origin does not. Use AI tools where they help, review and edit for accuracy and voice.
  • AI citations are volatile. 45.5% of AI Overview citations change between consecutive observations. Don't chase day-to-day citation movements; measure quarterly patterns at minimum.
  • Do not wait for a "3-month freshness cliff." The 3-month cliff is a myth in its absolute form. Freshness is a signal; quarterly updates reduce citation loss by ~3x. That's the real data.

Reference files

When planning content briefs or writing to the standard, read:

  • references/content-brief-template.md — Complete content brief template with field-by-field guidance and ski-ramp examples. Read when creating or reviewing a content brief.

Cross-skill handoffs

  • ← seo-keyword-research: Receive clustered keyword map and fan-out sub-queries as input.
  • → seo-onpage-optimization: Hand off content briefs for page-level ski ramp implementation.
  • → seo-link-building: Hand off information gain assets (original research, data studies) for distribution plans.
  • → aeo-ai-search-visibility: Hand off content for citation readiness review.
  • → seo-reporting: Hand off priority pages for tracking.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.