About this skill
AI Search (GEO) Playbook
A working playbook for getting cited, and then recommended, in AI answers: Google AI Overviews and AI Mode, ChatGPT, Perplexity, Gemini, Copilot and Claude. Your agent audits where you show up today, checks whether your key pages can be quoted on their own, checks robots.txt and CDN rules for AI crawlers, and turns the gaps into a ranked list of fixes.
What it covers. How each engine picks sources and where Google's official guidance differs from the other engines; the three pillars (structure, authority, presence); answer-block and FAQ patterns; llms.txt, /pricing.md and the newer agent-readable formats; crawler controls split by purpose (search discovery vs. user retrieval vs. training); and how to measure AI visibility when one run of a query proves nothing. Fourteen reference files go deeper on the hard questions: why you are cited but not shortlisted, wrong facts about your pricing in AI answers, LinkedIn and YouTube as citation sources, format shifts after model updates, attribution of AI traffic.
Why it is worth having. GEO advice goes stale in months. This skill dates its statistics, tells the agent to rank evidence (platform docs first, vendor studies last) and refuses to turn a study's percentage into your forecast. It also keeps Google's line that you should not write separate content for AI.
Pairs with. Our SEO Content Cluster Map plans topics for classic search; this playbook makes the pages you have quotable by AI engines.
Based on the ai-seo skill by Corey Haines (MIT): https://github.com/coreyhaines31/marketingskills
What we changed
- Renamed to
ai-search-geo-playbookand shortened the description (it was at the 1,024-character limit). - Added a "Deliverable" section so every run ends with the same four parts: visibility snapshot, top fixes, page rewrites, monitoring plan. It also tells the agent not to report AI-answer results it did not actually observe.
- Added attribution (
NOTICE.md, license text). All 14 reference files are unchanged. Full list inNOTICE.md.
Checked against
Claude Code 2.1.317 (desktop app, Claude Opus 5.5), 2026-11-07. Installed under .claude/skills/ai-search-geo-playbook/, loaded from SKILL.md and run end to end on a "why does ChatGPT recommend competitors" question about a not-yet-live site: routed to the citations-vs-recommendations and positioning references, ran the extractability check on the page, gave nine ordered fixes and an extractable hero rewrite. It did not query ChatGPT or Perplexity (no live access) and said so; the description and intake now say the same (NOTICE.md). The example output on this page is from that run, shortened. Not yet checked: automatic triggering from the skill description in a fresh session (the run started from an explicit request), and Codex, Cursor, GitHub Copilot and Gemini CLI, which read the same SKILL.md format but have not been run.
How to install
- Agents: copy the
ai-search-geo-playbookfolder to.agents/skills/ai-search-geo-playbook/(Codex, Cursor, Copilot, Gemini CLI) and/or.claude/skills/ai-search-geo-playbook/(Claude Code). - Zip: download the zip from this page and unpack it into one of those folders.
- ChatGPT / Claude: paste
SKILL.mdas the system prompt or project instructions, and attach the reference files that match your question (for examplereferences/agent-readiness.md).
Not affiliated with the upstream author or with any AI search platform named here.
