To run an AI search audit, build an inventory of 30 to 100 buyer prompts across discovery, comparison, and validation stages; test them across ChatGPT, Perplexity, Gemini, and Google AI Overviews; log brand mentions, competitor rank, cited sources, and factual errors; and verify robots.txt, schema, and HTML extractability before prioritizing fixes.
Step 1: Define Audit Scope and Buyer Prompt Inventory
Start with real commercial questions rather than isolated head keywords. A diagnostic audit for a single product needs 30 to 50 prompts, while a multi-product commercial baseline typically covers 80 to 150 prompts.
Group prompts by buying stage: problem awareness, category education, vendor discovery, head-to-head comparison, and procurement validation. High-intent discovery and comparison prompts reveal where revenue-influencing shortlists are won or lost.
- Record engine name, date, location, and exact prompt wording for every test run.
- Capture both the full answer text and every cited link or footnote.
Step 2: Map Cited Sources and Factual Accuracy
For each prompt, record whether your brand was absent, mentioned without a link, cited via an owned URL, or listed below competitors. Separately classify every cited domain into owned pages, review sites, editorial media, or community forums.
| Audit Symptom | Likely Root Cause | First Fix |
|---|---|---|
| Zero owned citations despite strong rankings | Bot blocks or JS-only content | Audit robots.txt and server-side HTML |
| Competitors cited on vendor lists | Missing comparison or spec pages | Publish structured comparison and criteria guides |
| Outdated pricing or wrong category | Stale third-party directories | Update G2, docs, and partner profiles |
Step 3: Verify Technical Readiness and Ship the Backlog
Check `robots.txt` for `OAI-SearchBot` and `PerplexityBot`, confirm canonical tags and XML sitemaps, and verify that core pricing, FAQ, and spec tables render in raw HTML without client-side scripts.
Use our AEO audit checklist for a printable worksheet and test live pages with the AI citation checker.
Frequently asked questions
How many prompts do I need for a reliable AI search audit?
Use 30 to 50 prompts for a focused single-product diagnostic check, and 80 to 150 prompts for a multi-persona commercial baseline.
What should an AI search audit deliver besides visibility scores?
Every row should connect a prompt, competitor shortlist, cited URLs, factual accuracy label, root-cause diagnosis, and a specific page or off-site owner to fix.
Do I need llms.txt to pass an AI search audit?
No. Focus first on crawlable HTML, clean sitemaps, accurate structured data, and third-party corroboration rather than experimental files.
Next steps & related pages
Want a baseline visibility check for your brand?
Save the CSV checklist above for manual audits, or request a baseline check on your target US buyer prompt while our automated tracking workspace rolls out.