What the engines do not tell you
ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews do not publish a ranking formula for businesses. Articles that list “the factors” are describing what tends to go together with being named, not a documented rule. We treat every such claim as a hypothesis until a measurement supports it for a given business and market.
What you can see: the answer itself
Ask a buyer question in an engine and read the answer. For each run we record whether the business is named, its position in the answer, the words used about it, the competitors named next to it and the sources the engine cites. The score is the share of runs in which you are named, per engine and per language, and every number links to the saved raw answer. The full method is on the method page.
Why one answer proves nothing
AI answers vary between runs. Asking once shows one possible answer; asking three times per question and per engine shows what is stable and what is chance. That is why the questions are written with the business before the first run and then frozen: a change in the result is then a change in the market, not in our wording.
What we work on, in the order an engine meets you
First the engine has to be able to read you: crawler access, text that is in the HTML without JavaScript, structured data, a sitemap. Then it has to recognise you: the same name, address, phone and description everywhere you are listed. Then it needs an answer to quote: pages that answer one buyer question each, directly in the first sentences. Last, it looks at what others say about you: articles, lists and directories. The eight steps are in the guide.
What you can check today
Open yourwebsite.com/robots.txt and look for the AI search crawlers; if one is blocked, you are out of those answers. The free check tests this and 14 more points on your homepage in about 20 seconds, without an account.