How to check if AI recommends your business

Ask the assistants the way a customer would, from the outside. Open ChatGPT, Gemini, and Perplexity while logged out, type the questions a stranger would ask about a business like yours in your area, and write down whether your name appears, whose name appears instead, and what the assistant says about you.

Why logged out, and why from the outside

The mistake almost everyone makes is testing from their own account, on their own network, where the engines already know them and bend the answer their way. That feels reassuring and tells you nothing. Your next customer arrives as a stranger, so you have to check as a stranger: logged out, ideally in a private window, asking the plain question they would ask.

If you only ever check from inside your own world, you are grading your own homework. The whole point is to see the answer your customer sees, not the flattering one the machine keeps for people it recognises.

The questions to actually ask

Do not ask for your business by name, because that proves nothing. Ask the category question, the one a customer types when they do not know you yet. “Best agriturismo near Cortona.” “Where to eat in Val di Chiana.” “Wine tasting near Arezzo without a car.” Then ask a few follow ups the way a real person would, narrowing by budget, by family, by season.

Run the same handful of questions across ChatGPT, Gemini, and Perplexity, because they do not answer alike and you want the pattern, not one lucky result. Keep the wording a customer would use, in the language your customer speaks, which for many Tuscan businesses means English as well as Italian.

What to write down

For each question, note three things. Did your name come up at all. If not, whose name did. And what did the assistant say, because sometimes you appear but the detail is wrong, an old address, the wrong hours, a service you no longer offer. A wrong mention can cost you more than no mention, and you only catch it by reading the answer carefully.

Do this across a page of questions and a picture forms fast. You will see the searches where you win, the ones where a specific competitor keeps taking your place, and the ones where the assistant simply has nothing to say about you, which is usually a sign it cannot read or trust your information.

Turning the check into a fix

A test is only useful if it changes something. When a competitor keeps winning a question, go read what the machine can see about them that it cannot see about you, usually a clearer page, a complete Google profile, consistent details, real reviews. When the assistant gets a fact wrong, trace it to the source, an old listing, a stale profile, and correct it there. When you are simply absent, the job is to give the machine something clear and trustworthy to find.

If you would rather not run all of this yourself, that is exactly the work I do in a snapshot: the same questions, across the same engines, recorded from the outside, handed to you as a short read. You can start with the free version below, and there is a fuller walk through in why AI often recommends the weaker business.