AI visibility audit checklist: 12 checks you can run yourself

By Rolf Kassel · Published 10 October 2026 · 7 min read

You can find out whether AI assistants recommend your company in an afternoon: write ten buyer prompts, run each on ChatGPT, Perplexity, Gemini and Google three times, and record who gets named, in what order, with which sources. This checklist walks through the twelve checks that matter.

Part 1: measure

1. Write ten prompts the way buyers type them. Two "best [category] for [segment]", two "[competitor] alternatives", two "[you] vs [competitor]", one pricing, one "is [you] good for [use case]", two trust ("best reviewed", "what to look for").
2. Run each prompt on four engines, three times. ChatGPT with search on, Perplexity, Gemini, Google (look for the AI Overview). Use a fresh chat each time.
3. Record four things per answer. Which vendors are named, in what order, the sentence used about you, and the cited URLs. A spreadsheet is enough.
4. Compute your mention rate. Answers naming you divided by all answers. Under 30% means you are mostly invisible.
5. Compute share of voice. Your mentions divided by all tracked vendors' mentions.
6. List the cited domains. Count how many answers cite each. The top five are your action list.

Part 2: diagnose

7. Check the listicles. Google your two "best X" prompts. Open the top five results. Are you on them? If not, that is why the engines do not name you.
8. Check your review-site profiles. On G2 and Capterra: review count, recency, category badge. Engines cite these pages constantly.
9. Check your entity consistency. Read your one-sentence description on your site, LinkedIn, Crunchbase and G2. If they differ, pick one.
10. Check for comparison content. Do you have "[you] vs [competitor]" and "[competitor] alternatives" pages? Do they have tables?
11. Check structured data. Paste your homepage into a schema validator. You want Organization with sameAs, and Product or Service with offers.
12. Check what the engines got wrong. Read the sentences from step 3. Wrong pricing, old names and stale positioning show up here, and each is fixable at the source.

What to do with the results

Rank the gaps by how many prompts they affect. Usually the order is: get on the cited listicles, build the comparison pages, refresh reviews, fix the entity description, then community threads. Re-run the same ten prompts in a month.

If you would rather have this done for you with 30 prompts and five engines, that is the free audit (10 prompts) or the full audit. The score we compute is documented on the AI Visibility Score page so you can reproduce it.