How do you do an AI visibility audit?

Question

Perciva · Accepted Answer

List 15–30 buyer questions across category, comparison, pricing, and trust; run each in ChatGPT, Perplexity, Gemini, and Claude using clean sessions; score every answer on four axes — mentioned, recommended, accurate, cited; log competitor appearances and wrong claims; then rank fixes by deal impact. A first audit takes one focused afternoon. Build the prompt list from real buyer language, not your messaging. Sources: sales-call questions, support tickets, the queries your comparison pages target, and the generic category question ('best [category] software'). Include prompts where you expect to win and prompts where you fear you lose — the audit's value is in the gap between expectation and reality. Run each prompt in each engine and record verbatim answers plus citations. Score simply: Were you mentioned? Were you the recommendation or just a list item? Is every claim about you correct? Are any of the cited pages yours? A spreadsheet with prompt × engine rows and those four columns is a complete audit artifact. Then triage. A wrong compliance claim on a trust prompt outranks a missing mention on a broad category prompt; displacement on a high-intent comparison outranks a mediocre description. Fix the top items, and decide your ongoing cadence — the audit is a snapshot, and answers will have moved within weeks.

How do you do an AI visibility audit?

See it for yourself