--- name: seo-aeo description: Optimize web content to rank in classic search (Google, Bing) and to be cited by AI answer engines (ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews). Use when drafting or auditing blog posts, landing pages, marketing site content, knowledge base articles, FAQ pages, or any public web copy meant to attract organic or AI-referred traffic. Activates on "SEO", "AEO", "GEO", "rank for", "optimize for search", "AI search", "answer engine", "blog post", "landing page", "meta description", "schema markup", "humanize for search". Skip for internal docs, decks, emails, or anything not published to the open web. --- # SEO / AEO Last reviewed: 2026-05-03 Produce content that wins in two ranking systems at once: traditional search (Google, Bing) and AI answer engines (ChatGPT, Perplexity, Claude, Gemini, AI Overviews). Roughly 38% of AI Overview citations come from pages already in the classic top 10, so the two are reinforcing, not separate. ## Core 12 Rules Ordered by leverage. Rules 1-4 are empirically validated by the GEO paper (Aggarwal et al., KDD 2024) — each shows ~30-40% relative lift in AI citation rates. The rest are practitioner consensus. 1. **Answer first, in 40-60 words, directly under the H1.** Inverted pyramid: claim, evidence, context. Wins featured snippets and AI extraction simultaneously. 2. **Add a statistic every 150-200 words, each with a named source and year.** Strongest in regulated, opinion, and explanatory topics. 3. **Quote a recognized authority verbatim, in quotation marks, with attribution.** Strongest in people, history, and explanatory topics. 4. **Cite external sources with inline links: at least 3 authoritative outbound links per article.** Perplexity in particular weights citations heavily. 5. **Use question-phrased H2s and answer in the first sentence below.** "How does X work?" / "What is Y?" wins People-Also-Ask, AI extraction, and FAQ schema all at once. 6. **Name the entity (brand, product, person) consistently and early.** Helps AI engines bind content to the correct entity. Pair with `Organization` or `Person` JSON-LD on every page. 7. **Author bylines with credentials and a linked author page** (with `Person` schema). E-E-A-T signal that carries to AI citation. 8. **Render content server-side or statically.** GPTBot, PerplexityBot, and ClaudeBot do not execute JavaScript. Client-only rendering is invisible to them. 9. **Show a visible "Last updated: YYYY-MM-DD" and update the `dateModified` field in schema.** Perplexity heavily weights freshness; YMYL topics get the same boost in classic search. 10. **Add JSON-LD: `Article`, `Organization`, and `FAQPage` where there's a real FAQ.** Skip `HowTo` (Google deprecated it Sept 2023). `FAQPage` rich results in Google are now restricted to government and healthcare, but the schema still helps AI citation. 11. **On-page baseline**: title tag 50-60 chars, meta description 150-160 chars, single unique H1, clean keyword-led slug, descriptive internal-link anchor text (not "click here"), 3-5 internal links per long-form piece. 12. **Strip AI-writing tells before publishing.** No "Additionally", "Furthermore", "delve", "landscape", "multifaceted", "testament to", em-dash overuse, copula avoidance ("serves as" → "is"). Run the `writing` skill's self-check pass after drafting. ## Structural Patterns That Help Both Engines - **One comparison table or numbered list per long-form piece.** Both Google snippet algorithms and LLM chunk extractors love them. - **Paragraphs ≤3 sentences, sentences ≤20 words on average, 8th-10th grade reading level.** Scannable for humans, cleanly chunkable for LLMs. - **Distinct, self-contained sections.** Each H2 should make sense in isolation, since AI engines often quote a single section out of context. - **Define jargon in the same sentence it's introduced.** AI extraction may surface that one sentence with no surrounding setup. ## Output Template for New Content When drafting from scratch, use this skeleton: ``` H1: Last updated: YYYY-MM-DD [40-60 word direct answer to the page's core question] ## [Direct one-sentence answer, then evidence with a stat + source.] ## [Direct one-sentence answer, then quote from a named authority.] ## [Table or numbered list.] ## FAQ [3-6 Q&A pairs covering related queries.] [JSON-LD block: Article + Organization + FAQPage] ``` ## Auditing Existing Content When asked to audit, score against the 12 rules and report: - **Pass / Fail per rule** with the specific gap. - **Top 3 fixes by leverage** (not by rule order). Almost always: rule 1 (answer-first), rule 2 (stats), rule 8 (SSR) when applicable. - **Quick wins under 30 minutes** vs. **structural rewrites**. - **Schema gaps**: list missing or malformed JSON-LD. Don't recommend keyword density targets, exact-match keyword stuffing, or `HowTo` schema. These are obsolete or counterproductive. ## Self-Check Before Publishing 1. Direct answer in first 40-60 words below H1? 2. At least one statistic per 150-200 words, each with source and year? 3. At least one named-authority quote with attribution? 4. 3+ authoritative outbound links? 5. H2s phrased as questions where the content allows? 6. `Article` + `Organization` JSON-LD present? `FAQPage` if there's a real FAQ? 7. Visible last-updated date matching `dateModified`? 8. Author byline with credentials and `Person` schema? 9. Title 50-60 chars, meta 150-160 chars, single H1, clean slug? 10. Page server-rendered (verify by curling without JS execution)? 11. Run the `writing` skill self-check to strip AI tells. 12. Read the first 60 words aloud. If a stranger could answer the page's core question from those 60 words, the inverted pyramid is right. ## References - GEO paper (Aggarwal et al., 2024): the source for rules 1-4's empirical claims. https://arxiv.org/abs/2311.09735 - Frase AEO Guide and totheweb GEO checklist: practitioner consensus that backs rules 5-11.