AI visibility audit: do ChatGPT, Grok and Gemini recommend your business?

Quick check: six questions, three assistants, two runs each, for $39, delivered in 2 working days. You receive the written result on WhatsApp or by email.

What the audit records, for every answer

Each answer from each assistant is read and written down the same way, so two answers, two assistants or two months can be compared line by line.

How we run it

1. Questions written like a customerBefore any test, we write the questions a buyer asks before choosing, in the buyer's language and with the city when the choice is local. Your business name is never included. Once set, the questions are not edited, so a later re-measure compares like with like.
2. Three assistants, two runs eachWe ask every question in the ChatGPT, Grok and Gemini web apps. Each run uses a new temporary or private chat with personalisation off. We ask each question twice because the same wording can return a different list.
3. Then we ask about you by nameWe then ask each assistant about your business by name: what it says it knows, which sources appear to support those claims, and whether each claim is accurate. An old address or wrong opening hours show up here.
4. Can they read your site at allWe check robots.txt rules for AI crawlers, the response each crawler receives, how much text is available without JavaScript, structured data, llms.txt and the sitemap.
5. Sources checked by handFor sources that appear across questions or assistants, we check how businesses are listed there and what the source requires. We do not buy placements, links or reviews.

What the report shows

Below is a real result, not a mock-up. We ran the same audit on our own agency site, in Spanish, for the Spanish market, before selling it to anyone.

The re-measure is due 60 to 90 days after the changes and uses the same six questions and settings, so the comparison is like for like.

Your report has the same parts:

• The results table, question by question, for each assistant and each run.
• Who is recommended when you are not, with the differences checked on their own pages.
• The sources that appear in answers for your category, and what it takes to be listed on each one.
• What the assistants say about you by name, each statement marked true or false.
• A plan in priority order, on your site and off it. Every item names the evidence behind it, who does it, and what we will check at the next measurement.
• A method note with dates, settings and the browser's country, so the test can be repeated.

The same shop, two very different answers

One run per assistant, by hand, September 2026.

Knitferatu is a knitwear shop we own. We asked ChatGPT and Grok the same eleven customer questions. ChatGPT cited the shop's own pages; Grok built its answers from marketplace listings, lists and gift guides, and never named the shop.

Those runs came before the current method existed, so read them as a pointer rather than a statistic. It is why the audit always covers three assistants and records where each answer comes from. Over 28 days the same shop logged 41 visits from ChatGPT and 21 from Google search.

Two audit sizes, plus a re-measure

What we will not do

We do not buy links, listings or reviews to move the answers, and we do not create fake accounts or posts dressed up as customers. A result holds for those questions, those assistants, those settings and those dates, and the report says so in writing.

When to start somewhere else

Sometimes AI answers are not where to start. If customers need someone immediately, like a plumber or a locksmith, the Google Business Profile may matter more than AI recommendations. If that is what the evidence suggests, the quick check says so.