AI Search Readiness (GEO) audit

Optimise for ChatGPT, Perplexity, and Gemini

/dashboard/sites/<id>/audit/ai-search measures how citable your client's content is to LLM-powered search engines — a separate ranking factor from classical Google SERP.

Composite GEO score (0-100)

Weighted by:

• 40% answer-extractability — Sonnet 4.6 extracts factual claims from page samples, scores extraction yield + claim density
• 25% citation pattern presence — heuristic detection of FAQ blocks, comparison tables, numbered lists, definition-then-example structure (the patterns LLMs love to quote)
• 15% AI-schema completeness — Article / FAQPage / HowTo / QAPage / Dataset JSON-LD hits
• 10% llms.txt + ai.txt — HEAD-fetched + validated. llms.txt is the emerging AI-first sitemap.
• 10% bot allowlist — robots.txt audit across 14 known AI bots (GPTBot, ClaudeBot, PerplexityBot, CCBot, Google-Extended, Bytespider, Applebot-Extended, Meta-ExternalAgent, etc) → recommend-by-intent (allow / block / partial)

Brand-mention-in-AI testing (Agency+)

Sends 5 probe queries to Anthropic, OpenAI, and Gemini APIs asking about the client's niche. We then detect:

• Whether the client's brand is mentioned at all
• Brand position (1st mention vs buried)
• Per-LLM cost-tracked

llms.txt + ai.txt

Two emerging community standards we validate against. llms.txt is the structured "what is this site, what content matters" file LLMs prefer to ingest. ai.txt is the bot-allowlist counterpart to robots.txt. Both are still draft-spec but Anthropic, OpenAI, and Google all honour them.