How do I find out if ChatGPT recommends my business?

You find out by asking ChatGPT the questions your customers would ask, without naming your business, and recording who it names. There’s no shortcut that I know of. I’m not aware of an official report that tells a local business how often it was recommended, so the evidence is what you collect yourself by running real prompts. Done carefully, it doesn’t take long and it tells you a lot.

The test, step by step

1. Write five to ten questions in a customer’s words. Each should include a service and a place and leave your business name out. For a hypothetical electrician in Vista: “Who should I call to install an EV charger in Vista?”
2. Start a new chat for every question, so that one answer doesn’t color the next.
3. Run each question several times, and on more than one day.
4. For every run, record the businesses named, the order they appear in, what is said about each, and any sources or links shown.
5. Only at the end, ask about your business by name, and check the facts it gives: services, service area, phone number and hours.

Everything after the first step depends on what prompts a local business should test, so ask the way a customer would.

Mistakes that give a misleading result

The most common one is asking about yourself by name. “Is [your company name] a good electrician?” will get you a description of your company, if it can find one. That shows what it knows about you. It doesn’t show whether it would bring you up to a stranger who has never heard of you, which is the thing you care about.

The second is testing from a chat or an account that already knows who you are. If earlier messages mention your company, or the product has remembered details about you, the answer may lean toward you in a way a customer’s wouldn’t. Products differ in what they remember and the settings change, so look for a temporary or signed-out mode and check how it currently works.

Stopping after one run is the other trap. Answers change from one try to the next, so a single reply, good or bad, is one sample.

Check whether it searched

Look for sources in the reply, because they tell you where the answer came from. An assistant can answer from what it learned in training or by searching the web at the time of the question. When it searches, the reply normally shows links. Open them. They are the pages it read, and seeing whether you appear on them is the most useful part of the whole exercise. A reply with no sources is more likely recall, which is slower to change and harder to influence.

How to read what you find

“Named in six of twenty runs, always behind the same two companies” is a finding you can act on. So is “never named, and the companies that are named all appear on one directory page where I’m not listed.” Each is a rate across many runs, which tells you more than any single answer. If a competitor keeps appearing, the usual reasons are the ones behind why ChatGPT recommends a competitor.

The test can’t tell you how many real customers asked those questions, or how many of your calls started there. And being named today doesn’t mean you’ll be named next month, since nobody can guarantee an AI recommendation.

Doing it once or keeping it going

One round of testing is enough to see whether there’s a problem. If you plan to change things and want to know whether the changes worked, you have to run the same questions the same way before and after. I do that as ChatGPT visibility work, and the ongoing version is AI visibility tracking.

The same test works on other assistants, and it’s worth running on more than one, because they don’t give the same answers.