# AI Search Extension Last reviewed: 2026-08-03. Recheck volatile provider sources in `data/seo-source-registry.json`. ## Google AI features Google states that AI Overviews and AI Mode use the same foundational SEO practices as Search. A supporting page must be indexed and eligible to appear with a snippet; there are no additional technical requirements, special AI schema types, or required AI text files. Use normal foundations: - crawl and index eligibility - important text available in rendered content - crawlable internal links - useful, reliable, people-first content - structured data matching visible content - page experience, Merchant Center, and Business Profile where relevant Google's newer generative-AI optimization guide documents a dedicated Search Console Generative AI performance report. Its broader AI-features documentation also describes AI-feature traffic as contributing to overall Search Console Web reporting. Treat these as different reporting views that may have rollout or property-availability conditions; verify the current property and documentation before claiming only one view exists. Google's 2026 guidance also says: - `llms.txt` neither helps nor harms Google Search visibility or ranking - no special Schema.org type is required for generative Search - there is no ideal AI-specific page length or mandatory “chunking” pattern - pages do not need to be rewritten for AI or every fan-out query - scaled pages created mainly to manipulate AI/search responses can violate spam policy Official sources: - https://developers.google.com/search/docs/appearance/ai-features - https://developers.google.com/search/docs/fundamentals/ai-optimization-guide ## Other AI search systems ChatGPT Search, Perplexity, Copilot, Gemini surfaces, and other systems may differ by provider, product, model, geography, login state, and date. Verify current official crawler/search documentation and observed source behavior per provider. Use [Engine Matrix](engine-matrix.md). For OpenAI, keep these controls separate: - `OAI-SearchBot`: automatic Search discovery/surfacing - `GPTBot`: content that may be used for foundation-model training - `ChatGPT-User`: user-triggered visits that are not automatic Search crawling and may not follow robots rules identically For Perplexity, current documentation distinguishes `PerplexityBot` for search from other provider behavior. Verify published IP ranges when bot identity matters. Do not assume: - one robots directive controls every provider - a training crawler is the same as a search/citation crawler - one citation test predicts future answers - a mention means recommendation - structured data or `llms.txt` guarantees retrieval or citation ## Evidence ladder Track separate stages: 1. `eligible`: content is accessible under the provider's documented conditions 2. `retrieved`: the system appears to use or surface the source 3. `cited`: the source receives an explicit citation/link 4. `mentioned`: the entity or brand is named 5. `recommended`: the entity is included in a preferred shortlist 6. `converted`: measurable downstream user action occurs Never collapse these stages into “AI visibility improved.” ## Observation protocol For a useful AI-search check, record: - query set and selection rationale - provider and product surface - date/time, market/language, device, and login state when relevant - answer, citations, cited URL, mention framing, and recommendation state - repeated observations or independent reviewers when claims matter - referral/conversion evidence when available Treat manual prompt checks as sampled observations, not a stable ranking report. ## Optional agent-readable files An `llms.txt`, markdown knowledge bundle, or simplified pricing/specification file may improve documentation or agent operability in some contexts. Before proposing one: - identify the consumer and documented behavior - avoid duplicating or exposing private content - keep it consistent with visible canonical pages - define how freshness will be maintained - label ranking/citation impact as `missing evidence` unless measured Do not add a file solely because an SEO checklist claims it is required.