Skip to content

geo-audit

Analyze any website for Generative Engine Optimization: how well AI platforms can discover, understand, and cite the site's content. Generates scored reports with prioritized recommendations for improving AI citation rates.

  • Area: Marketing
  • Category: marketing seo
  • Level: beginner
  • Input: url
  • Output: document
  • Setup: none
  • Roles: Head of Marketing

Performs Generative Engine Optimization audits for AI citation across ChatGPT, Perplexity, Gemini, Bing Copilot, and Google AI Overviews.

How you ask

  • “Run a GEO audit on our website for AI visibility”
  • “Check how well our site is optimized for AI citations”
  • “Audit our brand's presence in ChatGPT and Perplexity results”
  • “Generate an llms.txt file for our domain”

Inputs

  • marketing_context — Positioning and messaging context.
  • url — Page or site to assess for AI answer-engine visibility
  • url — Root domain to assess
  • buyer_queries — Buyer-intent queries to test citation against
  • url — Site to assess
  • platform — The AI answer engine to audit citation visibility on — google-ai-overviews, chatgpt, perplexity, gemini, or copilot; all produces one scorecard per platform plus a cross-platform summary. (google-ai-overviews | chatgpt | perplexity | gemini | copilot | all)
  • url — Site to generate llms.txt for
  • include_paths — Sections to expose; defaults to public content only
  • panel_config — Versioned prompt panel; built on first run if absent
  • url — Brand domain being tracked
  • platforms — Platforms to run; defaults to the panel config
  • runs_per_prompt — Runs per prompt per platform; n>=3, defaults to the panel config
  • business_name — Exact canonical business name, used verbatim as question 5
  • what_they_sell — What the business sells, in the owner's words — the category questions are built from this
  • location — Primary town or region; omit for a pure-online business
  • top_competitor — The competitor the owner names first; seeds the alternatives question
  • platforms — Platforms to ask; defaults to ChatGPT and Perplexity

Outputs

  • One probe per prompt per platform per repeat, typed geo-prompt-probe. Never hand-edited — a probe records that a question was asked and answered.
  • Ten named/not-named results plus the single highest-impact fix
  • geo-audit-pdf (pdf)
  • geo-audit-report (markdown)
  • llms-txt (markdown)
  • llms-txt-client (markdown)
  • prompt-panel-pdf (pdf)
  • prompt-panel-report (markdown)
  • {output_base}andare the GEO folder of the scope's marketing area (/or/); with no vault contextis` and the prompt panel does not run. Full layout: references/deliverables_and_reporting.md