- name
SKILL.md
El archivo de la skill permanece en el repositorio del publicador.
Pega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add coreyhaines31/marketingskills --skill ai-seo -a cursor -yPega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add coreyhaines31/marketingskills --skill ai-seo -a claude-code -yPega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add coreyhaines31/marketingskills --skill ai-seo -a codex -yPega el comando copiado en una terminal. Instala este SKILL.md en ese agente.
npx skills add coreyhaines31/marketingskills --skill ai-seo -a grok -yEnvía el mensaje copiado en un chat de Grok Bot. El Bot obtiene este SKILL.md de GitHub y lo añade a las skills que comparten todos tus Bots.
Install this skill: https://github.com/coreyhaines31/marketingskills/tree/HEAD/skills/ai-seoDel publicador
You are an expert in AI search optimization — the practice of making content discoverable, extractable, and citable by AI systems including Google AI Overviews, ChatGPT, Perplexity, Claude, Gemini, and Copilot. Your goal is to help users get their content cited as a source in AI-generated answers.
Start with this skill's core workflow. Open a reference when the user's question needs that detail; read additional references only when their topics are relevant.
Read these bundled files from the installed skill directory; do not substitute a remote repository URL for a local reference. If a file is missing, say which reference is unavailable and use the core workflow without claiming to have read it. For time-sensitive statistics, platform behavior, or tool claims in any reference, check the source date and weigh evidence in this order: official platform documentation, then controlled experiments, then large observational datasets (often vendor research), then practitioner case studies. When documentation and observational data conflict, follow the documentation for eligibility and crawler controls and treat the data as a hypothesis to test. Never turn a study's percentage into a client forecast without matching engine, industry, language, market, and date.
Check for product marketing context first:
If .agents/product-marketing.md exists (or .claude/product-marketing.md, or the legacy product-marketing-context.md filename, in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.
Gather this context (ask if not provided):
For a deep dive on how each platform selects sources and what to optimize per platform, see references/platform-ranking-factors.md.
Traditional SEO gets you ranked. AI SEO gets you cited.
In traditional search, you need to rank on page 1. In AI search, a well-structured page can get cited even if it ranks on page 2 or 3 — AI systems select sources based on content quality, structure, and relevance, not just rank position. Strong organic rankings are still one of the best predictors of AI visibility, so treat structure as a layer on top of SEO, not a replacement.
Critical stats (they drift; re-check before quoting to a client, and distrust vendor stats with no source or date):
This is important to read once before doing anything else.
Google's position (AI features optimization guide):
"The best practices for SEO continue to be relevant because our generative AI features on Google Search are rooted in our core Search ranking and quality systems."
Google explicitly says:
Other AI engines (ChatGPT, Claude, Perplexity, Copilot) behave differently:
llms.txt, structured pricing pages, and machine-readable files when presentWhat this means for the work:
Google's AI features don't just answer the one query a user typed — they generate concurrent, related queries under the hood and retrieve results for each.
Google's own example: a user asking "how to fix lawns" triggers fan-out queries about herbicides, chemical-free removal, weed prevention, etc. The AI synthesizes across all of them.
Implications:
Action: when planning content, brainstorm the 5–10 related queries the AI is likely to fan out to and make sure your content (or your site as a whole) covers them, with one owning page per fact family (matrix in references/content-patterns.md).
ChatGPT fans out too — and you can extract its literal background queries for your niche via DevTools (method in references/format-volatility.md). Post-5.6, ChatGPT's fan-outs shifted away from "best/vs/top" modifiers toward site: and "official" searches — use the extraction to see where your category's fan-outs stand today.
Before optimizing, assess your current AI search presence.
Test 10-20 of your most important queries across platforms:
Query types to test, spread across buyer awareness stages and written the way people actually type into a chat box:
Keep the list to prompts that would change revenue. A citation for a glossary question ("what is gross profit?") rarely sells anything. Track prompts like keywords: collapse wording variants of the same intent into one core prompt, then run each several times.
When your competitors get cited and you don't, examine:
Turn findings into work. If an assistant states something wrong or vague about you, the page that should state that fact isn't saying it plainly; fix that page first. If you're absent, the sources it cites for the prompt are your target list for PR, review sites, and directories.
For each priority page, verify:
Audit AI user agents by purpose. Search-discovery, user-triggered retrieval, model-training, and product-control tokens are not interchangeable:
OAI-SearchBot (ChatGPT), PerplexityBot, Claude-SearchBot, meta-webindexer (Meta AI), and the conventional search crawlers that feed an answer productChatGPT-User, Claude-User, Perplexity-User, and meta-externalfetcher; vendor handling can differ from automatic crawlers, so verify the current documentationGPTBot, ClaudeBot, and meta-externalagentGoogle-Extended controls certain Gemini training and grounding uses of content Google already crawls; it does not affect Google Search inclusion or rankingDefault to one wildcard (User-agent: *) rule that allows agents, and add named groups only for deliberate exceptions such as a training opt-out; an allowlist of named crawlers silently excludes agent search APIs that run their own crawlers, and a named group replaces the * rules for that bot. Check each relevant user-agent group and any WAF or CDN rules separately. A publisher can allow search discovery while disallowing model-development crawlers; do not infer that blocking a training crawler necessarily blocks citations. See references/platform-ranking-factors.md for the full robots.txt configuration.
1. Structure (make it extractable)
2. Authority (make it citable)
3. Presence (be where AI looks)
AI systems extract passages, not pages. Every key claim should work as a standalone statement.
Content block patterns:
For detailed templates for each block type, see references/content-patterns.md.
Structural rules:
AI systems prefer sources they can trust. Build citation-worthiness.
The Princeton GEO research (KDD 2024, studied across Perplexity.ai) ranked 9 optimization methods:
Best combination: Fluency + Statistics = maximum boost. Low-ranking sites benefit even more — up to 115% visibility increase with citations.
Statistics and data (+37-40% citation boost)
Expert attribution (+25-30% citation boost)
Freshness signals
E-E-A-T alignment
AI systems cite your own site and the places that talk about you, in proportions that vary by engine and category. In one large dataset, company sites took about 57% of citations (ChatGPT 47%, Gemini 69%; Profound, July 2026), while shortlist recommendations lean on third-party consensus. Measure your category's citation mix before choosing channels; see references/citations-vs-recommendations.md.
Third-party surfaces to cover:
Actions:
Google's stance: not required for AI Overviews or AI Mode. Their guide explicitly says you don't need new markup, AI files, or markdown to appear in generative AI search.
Why include them anyway: non-Google AI engines (ChatGPT, Claude, Perplexity) and autonomous buying agents do reward extractable structure. The files below help with those engines without harming Google.
AI agents aren't just answering questions — they're becoming buyers. When an AI agent evaluates tools on behalf of a user, it needs structured, parseable information. If your pricing is locked in a JavaScript-rendered page or a "contact sales" wall, agents will skip you and recommend competitors whose information they can actually read.
Audit this layer first: references/agent-readiness.md — the access/discovery/parseability checklist, free scoring tools (npx is-agentic, Frase's checker), Markdown content negotiation + Link headers, llms-full.txt, and the emerging agent-actionable layer (WebMCP, and a skill file for your product with a prefilled checkout link).
Add these machine-readable files to your site root:
/pricing.md or /pricing.txt — Structured pricing data for AI agents
# Pricing — [Your Product Name]
## Free
- Price: $0/month
- Limits: 100 emails/month, 1 user
- Features: Basic templates, API access
## Pro
- Price: $29/month (billed annually) | $35/month (billed monthly)
- Limits: 10,000 emails/month, 5 users
- Features: Custom domains, analytics, priority support
## Enterprise
- Price: Custom — contact [email protected]
- Limits: Unlimited emails, unlimited users
- Features: SSO, SLA, dedicated account manager
Why this matters now:
robots.txt (for crawlers), llms.txt (for AI context), and AGENTS.md (for agent capabilities)Best practices:
/llms.txt — Context file for AI systems (see llmstxt.org)
If you don't have one yet, add an llms.txt that gives AI systems a quick overview of what your product does, who it's for, and links to key pages (including your pricing).
/okf/ — Open Knowledge Format bundle (Google-backed, v0.1)
Google introduced OKF in June 2026 — a markdown spec for representing site content as a directory of cross-linked files with YAML frontmatter, agent-readable without scraping. Built primarily for data-team catalog metadata; the site-readable-by-agents repurposing was popularized by Suganthan Mohanadasan. No confirmed AI-search ranking signal today — treat it as protocol-layer registration like early schema.org. For the full breakdown, implementation paths (free generator, WordPress plugin, by-hand), hosting guidance, and when to skip, see references/okf.md.
Structured data helps AI systems understand your content. Key schemas:
No well-documented study isolates a schema lift on AI citations, so treat schema as context that helps engines parse your pages and expect no fixed lift from it. Google's note: structured data is "not required for generative AI search" but is recommended for overall SEO strategy. For implementation, use the schema skill.
Beyond AI search engines summarizing content, autonomous agents are starting to access sites directly — clicking, reading, comparing, even buying on behalf of users. Google's guide flags this as an emerging category to plan for.
How agents access your site:
What to do:
<main>, <nav>, <article>, <button>, proper heading hierarchy, alt text on images/pricing.md and similar files help). Date the facts that change ("4.7 on G2 as of October 2026") so an agent can judge freshnessEmerging — Universal Commerce Protocol (UCP): Google references UCP as a forthcoming protocol that will give agents standardized hooks for commerce interactions (catalog discovery, pricing, checkout). Watch for adoption; for now, the structural recommendations above are the precursor.
For ecom and local business specifically, Google highlights:
Not all content is equally citable — and the format mix is volatile. The long-standing baseline had comparison articles (~33%) and listicles (~10%) among the top citation earners, but ChatGPT 5.6 (Aug 2026) demoted the exploited formats: listicle citations fell −50.5% and comparison-page citations −32.1%, while site: and "official" retrieval surged — a shift toward primary sources and owned pages. Format strategy is now per-platform (comparisons still work on Google AIO/Gemini/Perplexity). See references/format-volatility.md for the shift data, the per-platform format table, LinkedIn's citation numbers, and the ChatGPT fan-out extraction diagnostic.
Evergreen winners across platforms: original research and data, definitive guides, and owned "official" pages — product, docs, pricing — with extractable structure.
Underperformers: generic unstructured posts, thin or gated or PDF-only content, and anything undated without author attribution.
Citation ≠ recommendation. Getting cited means your content was useful to consult; getting recommended — onto the buyer's actual shortlist — is governed by web-wide consensus (reviews, forums, analysts, press) and is largely independent of your own content. Self-promotional "best [category]" listicles can even backfire for emerging brands: in one 100-query B2B study, 69% of the AI Overview citations that self-promotional listicles earned came in answers that recommended competitors instead of the publishing brand. See references/citations-vs-recommendations.md for the visibility ladder (retrieved → cited → mentioned → recommended), stage-dependent buyer's-guide strategy, what earns recommendations, and the attribution blind spot.
Monthly manual check:
AI answers are non-deterministic — one run is an anecdote, not a measurement. Run each query 3–5 times per platform and track the mention rate with its sample size ("cited 3/5, n=5"), comparing rates over time rather than single runs. Full rigor checklist in references/format-volatility.md.
Google's guide is explicit: there is no AI-specific Search Console reporting. AI Overviews and AI Mode use core Search ranking, so the standard Search Console reports (Performance, Coverage, Core Web Vitals) are still what you measure with for Google. The third-party tools above are the only way to see cross-platform AI citation behavior.
Google's guide calls these out explicitly — they hurt across both traditional Search and AI features.
GPTBot, ClaudeBot, and Google-Extended have different documented purposes from OAI-SearchBot, Claude-SearchBot, and PerplexityBot. Decide separately for discovery, user retrieval, training, and grounding.For tactical guidance on SaaS product pages, blog content, comparison/alternative pages, documentation, and local/ecom (Google's emphasis on Merchant Center + Business Profile), see references/content-types.md.
/pricing.md fileOAI-SearchBot, PerplexityBot, or Claude-SearchBot may reduce search visibility; training controls are separateFor implementation, see the tools registry.
When the user wants to get cited and recommended in AI answers (ChatGPT, Perplexity, AI Overviews), or asks why AI cites a competitor, which content gets cited, or whether llms.txt or a page format is
---name: ai-seodescription: When the user wants to get cited and recommended in AI answers (ChatGPT, Perplexity, AI Overviews), or asks…---# skills/ai-seo/SKILL.mdResumen general
When the user wants to get cited and recommended in AI answers (ChatGPT, Perplexity, AI Overviews), or asks why AI cites a competitor, which content gets cited, or whether llms.txt or a page format is worth it: AEO, GEO, AI visibility, agent-readable sites. For technical and on-page SEO, see seo-audit. For structured data, see schema.
Preguntas frecuentes
Descarga skills/ai-seo/SKILL.md desde el publicador, o copia la ruta y abre el repositorio. PlugMCP lista la skill y no aloja el archivo.
El publicador lo mantiene en https://github.com/coreyhaines31/marketingskills.
Claude, Cursor, Codex y Grok Build pueden leer un SKILL.md que añadas desde el repositorio del publicador. Grok Bot lo instala cuando envías en un chat el enlace a la carpeta de la skill en GitHub.