aeo

# Answer Engine Optimization (AEO) AEO is the practice of structuring and evidencing content so AI systems (ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews/AI Mode, Copilot) select and cite it as the authoritative answer, rather than optimizing purely for click-through search rankings. This skill distills the consensus practices confirmed across the top AEO-related skills on skills.sh, and flags the few points where sources disagree or rely on unverified specifics. See `references/audit-methodology.md` for how this skill was compiled. ## AEO vs SEO | | SEO | AEO | |---|---|---| | Goal | Rank on page 1, drive clicks | Be cited as the answer | | Audience | Humans scanning results | LLMs generating responses | | Success metric | Position 1-10, CTR | Citation count, share of answer | | Content shape | Often longer, keyword-aware | Concise, self-contained, fact-dense | | Links | Backlinks for authority | Citations to primary sources | They overlap heavily and reward the same underlying quality signals (E-E-A-T, HTTPS, real content). Don't treat them as competing investments — strong SEO fundamentals are close to a prerequisite for AEO, not a substitute for it. ## When to use this skill - Auditing existing content for AI-citation readiness - Writing new content intended to be quoted by AI assistants - Implementing structured data (JSON-LD) for AI extraction - Deciding robots.txt policy for AI crawlers - Building or updating `llms.txt` - Explaining to a stakeholder why "SEO is fine but we're invisible in ChatGPT" ## When NOT to use this skill - Pure click-through SEO with no AI-citation intent — use general SEO practices instead - Brand-voice content with no factual claims (there's nothing to cite) - Breaking news / time-sensitive content — LLM training and crawl lag means citations arrive weeks to months later, if at all - Topics where LLMs already have strong training-data coverage (e.g. elementary math) — little citation upside ## Core principles (high confidence, cross-validated) These are consistent across every AEO resource surveyed for this skill, with no disagreement: 1. **Answer first, explain after.** Open sections with a direct 1-3 sentence answer to the question the section addresses. AI extracts the first clear statement it finds; buried conclusions don't get cited. 2. **Write self-contained, quotable claims.** Each key sentence should make sense with zero surrounding context — roughly 15-20 words, naming the actual subject instead of "it" or "this." 3. **Structure over prose.** Tables, numbered steps, and bullet lists get parsed and extracted far more reliably than paragraphs. See `references/extractable-content-patterns.md` for copy-ready templates. 4. **Every real claim needs a citation.** Named source, a specific number, and a date. Vague hedges ("may help," "many believe") get filtered out, not cited. 5. **E-E-A-T is the shared trust rubric.** Experience, Expertise, Authoritativeness, Trustworthiness — Google's own framework, and it governs AI citation selection too. See `references/eeat.md`. 6. **Structured data matters, `FAQPage` most of all.** JSON-LD schema (`FAQPage`, `Article`+author, `HowTo`, `Organization`) is consistently ranked as the highest-leverage technical investment. See `references/structured-data.md`. 7. **Freshness is real but must be substantive.** Visible `datePublished`/`dateModified` and genuine periodic updates matter. Changing a date without meaningful edits doesn't help and can hurt trust. 8. **A blocked AI crawler zeros out that platform.** Before any content work, confirm `robots.txt` isn't blocking the AI bots you want visibility with. This is a five-minute check with outsized impact. See `references/crawler-access.md` for the current bot matrix and where sources disagree. 9. **Single-topic pages beat sprawling guides.** A focused page on one concept extracts and gets cited more reliably than a comprehensive page covering ten things. ## The one credible number to actually cite Aggarwal et al., "GEO: Generative Engine Optimization" (KDD 2024) is the closest thing to a peer-reviewed empirical source in this space, and it's the one several independently-authored AEO skills converge on: | Method | Visibility boost | |---|---:| | Cite sources | +30-40% | | Add statistics | +37-40% | | Add quotations | +30% | | Authoritative tone | +25% | | Improve clarity/fluency | +15-30% | | Keyword stuffing | -10% (actively harmful) | Low-authority ("challenger") sites benefit disproportionately, up to 115% visibility increase; already-top-ranked sites that over-optimize can lose visibility. Match optimization intensity to existing authority — see `references/audit-checklist.md`. Treat every other specific percentage you encounter in AEO content ("58% of queries are zero-click," "47% of searches show AI Overviews," specific 2026 product names and rollout dates) as directionally useful but unverified. Check primary sources before repeating them as facts. ## Workflow 1. **Check crawler access first.** Read `references/crawler-access.md`, verify `robots.txt` against the current bot matrix for the platforms you care about. Fix this before anything else. 2. **Audit existing content.** Score against `references/audit-checklist.md` (E-E-A-T, structure, citation density, schema, freshness). 3. **Apply extractable content patterns.** Use `references/extractable-content-patterns.md` templates for definitions, FAQs, comparison tables, and steps. 4. **Add structured data.** Implement JSON-LD per `references/structured-data.md`, validate with Google's Rich Results Test and the Schema.org Validator. 5. **Consider entity and access signals.** `llms.txt`, Wikipedia/Wikidata presence, consistent NAP — see `references/entity-and-llms-txt.md`. 6. **Measure.** Manual test priority queries across platforms monthly, cross-reference with Google Search Console's AI Overview/AI Mode impressions where available, track referral traffic from `chatgpt.com`, `claude.ai`, `perplexity.ai`. See `references/measurement.md`. 7. **Re-audit quarterly.** AI products change faster than search algorithms historically did; treat any AEO snapshot as perishable. ## References - `references/eeat.md` — E-E-A-T pillars, implementation, YMYL considerations - `references/extractable-content-patterns.md` — definition, FAQ, comparison table, steps, statistic, and quote block templates - `references/structured-data.md` — JSON-LD patterns for Article, FAQPage, HowTo, Organization, Product, BreadcrumbList - `references/crawler-access.md` — AI bot matrix, robots.txt policy, and where sources disagree - `references/entity-and-llms-txt.md` — llms.txt, entity/knowledge-graph signals, NAP consistency - `references/platform-behavior.md` — how ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews/AI Mode differ in source selection - `references/measurement.md` — what to track, which tools exist, how to diagnose a citation drop - `references/audit-checklist.md` — scoring rubric and authority-aware optimization strategy - `references/audit-methodology.md` — how this skill was compiled, and its confidence levels ## Anti-patterns - Keyword stuffing — helps nothing in AEO, actively measured as harmful (-10% visibility) - Generic listicles with no original insight - Vague, hedged language ("may help," "could potentially," "many people") - FAQ answers over ~50-100 words, or buried conclusions - Manufacturing FAQ schema on pages with no genuine FAQ content — this gets penalized, not rewarded - Optimizing for one AI platform's quirks at the expense of shared E-E-A-T fundamentals — citation behavior correlates highly across platforms - Publishing pure LLM output with no human review — low-distinctiveness text gets deprioritized by retrieval systems that are specifically looking for a distinctive signal - Over-optimizing an already-authoritative page — established sites should use a light touch