How to Check if AI Recommends Your Business in 7 Minutes

The manual version takes about seven minutes and costs nothing. You open each AI engine, ask the question a customer in your town would ask, and write down whether your business is one of the sources the answer uses. Do it once and you have a snapshot. Do it eight times per engine and you have something you can actually trust.

Key takeaways

Ask category questions like "best mobile dog groomer in Sacramento", not "tell me about Paws on Wheels". A recent Foundation and AirOps analysis of 28 million AI responses found 91 to 97% of AI citations come from category queries rather than brand searches.
Check ChatGPT, Google's Gemini or AI Mode, Perplexity and Claude separately. They pull from different places, so a business named in one can be missing from the next.
If your business is not among the sources the answer used, do not describe that as "AI could not find me". You are seeing what the answer cited, not what the engine considered.
Across 9,454 question-by-business observations in our research, 84.3% of the checks came back with the business not among the sources the answer used. Missing the answer is the norm, not the exception.
The single strongest predictor of appearing was Google organic presence for that same question (odds ratio 12.74). We cannot tell you that fixing one causes the other, only that they travel together.
One check per engine tells you almost nothing. Noise from a single answer is why any serious test asks the same question several ways and looks at the pattern.

The 7-minute manual check

Set a timer, open four browser tabs, and use the same three questions in each.

Tab 1: ChatGPT. Ask "who is the best mobile dog groomer in [your town]?" Then ask "I need a mobile dog groomer near [your zip code] this week, who should I call?" Then ask "mobile dog grooming for a senior dog in [your town], any recommendations?" Write down every business the answer names, and the sources listed at the bottom.

Tab 2: Google Gemini (or the AI Mode in Google Search). Same three questions. Same notes.

Tab 3: Perplexity. Same three questions. Perplexity shows sources most clearly, so pay attention to which sites it pulled from.

Tab 4: Claude. Same three questions.

Now look at your notes. For each engine, count how many of the three answers included your business as one of the sources it used to build the answer. Zero out of three is a signal. One out of three is noise. Three out of three is the outcome you want.

You will notice something almost immediately: the same three questions produce different names in different engines. That is not a bug. ChatGPT leans on Foursquare Places and now Yelp, falling back to Bing Places. Gemini is grounded in Google Maps. Perplexity leans on Yelp, TripAdvisor and industry directories. Claude leans on Brave Search. If your business is strong on Yelp but thin on Google Maps, you can be named all day in Perplexity and never once in Gemini.

What the answer means when you are not in it

If your business is not among the sources an answer used, resist the urge to describe that as the engine "not finding you". You cannot see what an engine retrieved and then discarded. You can only see the bibliography it printed at the bottom of the answer. Everything else is guessing.

What we do know, from our own research, is that missing the answer is the normal outcome. We ran 9,454 question-by-business observations for local businesses in the US. In 84.3% of them, the business was not among the sources the answer used. So 94% of the observed failure happens before the answer is even written. The question is not "did the engine like my page". The question is "did the engine ever consider my page as a source for this question".

The strongest signal we found for clearing that stage was one thing: whether the business showed up in ordinary Google search results for that same question. The odds ratio was 12.74, meaning the difference between showing up in Google for the query and not showing up dwarfed everything else we measured. We tested nine technical page signals as alternatives (the kind of things audit tools grade you on) and none of them survived a strict statistical correction.

We want to be careful about what that means. It is an association, not a proven cause. We cannot tell you that fixing your Google presence will make ChatGPT recommend you. Anyone who tells you they can prove that is guessing. What we can say is that being visible to Google for the question a customer asks is the biggest measured lever in a system that has very few measurable levers. If you were missing from the answers you just checked, that is where the honest attention goes first.

Ask the questions a customer would actually ask

The most common mistake on a manual check is asking "what do you know about [your business name]?". That question always produces something, because you gave the engine your name. What you learn from it is close to nothing.

The question that matters is the one a customer types when they do not know you yet. For a mobile dog groomer, that looks like:

"best mobile dog groomer in Sacramento"
"mobile dog grooming near me for a nervous rescue"
"who does dog grooming at your house in East Sac"
"mobile groomer for a doodle in Sacramento"
"cheap mobile dog groomer Sacramento"

Notice that only one of those is generic. The rest have a real customer detail attached: the dog's temperament, the neighborhood, the breed, the budget. This is closer to what people actually ask, and it is closer to what AI answers actually get built around. The Foundation and AirOps analysis of 28 million AI responses in Q4 2025 found that 91 to 97% of AI citations come from these kinds of category queries, not from brand searches.

If you are a plumber, run the same drill with "emergency plumber [town]", "tankless water heater install [town]", "plumber for a slab leak [town]". If you are a dentist, run "family dentist accepting new patients [town]", "dentist for a kid with autism [town]", "same-day crown [town]". The specificity is the point. Your customers are specific.

The four engines pull from four different places

This is the part almost every local business owner gets wrong. There is no single AI recommendation system. There are at least four, and they source local business names from different places.

ChatGPT. OpenAI signed a partnership with Foursquare on December 16, 2024, covering the Places API and its database of 100M+ points of interest across 200+ countries. Then on July 23, 2026, OpenAI announced a licensing deal with Yelp, bringing Yelp reviews, ratings and photos into ChatGPT. Reservations and waitlist support followed on August 10, 2026. If your business is thin on Foursquare and Yelp, you are working uphill in ChatGPT.

Gemini and Google's AI Mode. These are grounded in Google's own local index, which means your Google Business Profile carries almost all of the weight. If your profile is unclaimed, has the wrong hours, or is missing service categories, this is where you feel it first.

Perplexity. Yelp is dominant here. That Foundation and AirOps analysis of 28 million AI responses found Yelp took 512,680 citations, 3.4 times the next platform. Perplexity also pulls from TripAdvisor and industry-specific directories more than the others.

Claude. Claude uses Brave Search as its grounding layer, which behaves closer to a general web index than to a local database. What Brave sees for your town, Claude sees.

If you want to go deeper on this, where each AI actually gets local business names walks through it engine by engine.

Doing this check by hand once is barely enough

Here is the honest limit of the seven-minute version. One answer from one engine is one data point. Answers are noisy. The same question on the same engine can name different businesses on different days, at different times, from different devices, in different browsers. If you ran your check on a Tuesday afternoon and your business was not named, you learned that on one Tuesday afternoon it was not named. You did not learn that "AI does not recommend you".

That is why a serious check does not ask three questions per engine. It asks eight. And it does not ask them once. It asks them repeatedly and compares the answers against each other, so one weird result does not decide the whole picture.

You can do that yourself. Set a calendar reminder for once a week, keep a spreadsheet, and after a month you will have thirty-two observations per engine. That is enough to tell signal from noise. Or you can run the free AI Visibility Check, which does the eight-questions-per-engine version in the background and hands you the pattern instead of the raw notes. Both work. The point is not the tool. The point is that a single check on a single afternoon is not the same thing as an answer.

If the pattern comes back with your business missing more often than not, why AI names a competitor in your town instead of you is the next honest read. And if you want to understand the two very different things the word "citation" means in this world (a directory listing and an AI answer both use the same word for different things), we wrote local citations vs AI citations for exactly that confusion.

What to write down while you check

Keep it simple. For each of the three questions in each of the four engines, note four things:

Did the answer name your business at all? Yes or no.
Were you one of the sources listed at the bottom? Yes or no. This matters more than being mentioned in passing.
Which competitors were named? Write down every business that got a mention. This is the roster you are actually competing against for that question.
Which sources did the answer pull from? Google Business Profile, Yelp, a local best-of list, a news article, a Reddit thread, a directory. If the same source shows up in every answer, that source matters.

By the time you have done this for four engines and three questions, you will have 12 rows in a notebook and a clearer picture of your local AI position than most of your competitors have ever gathered. What you do with it is a separate question, and it starts with your Google presence for the exact phrases that came back empty.

Frequently asked questions

How often should I check if AI recommends my business?

Monthly is enough for most local businesses. Weekly if you are actively working on your Google Business Profile, reviews, or getting named in local best-of roundups, because you want to catch when the pattern changes. Daily is too often and mostly measures noise.

What if my business is named in one engine but not the others?

That is normal and it tells you where your gap is. If you are strong in Perplexity but weak in Gemini, your Yelp presence is doing work your Google Business Profile is not. If you are strong in Gemini but weak in ChatGPT, the reverse. This is one reason checking whether AI recommends your business is worth doing across engines, not just one.

Does asking ChatGPT about my business by name count?

No. When you type your own business name, the engine has almost no choice but to answer with something about you. That teaches you nothing about whether a new customer, who has never heard of you, would find you. Always ask the category question a stranger would ask.

Why does the same question give different answers each time?

Because AI answers are generated fresh, and they draw from ranked pools of possible sources that shift with tiny changes in wording, timing and the engine's own updates. This is exactly why one manual check is not enough and why real testing runs the same question several times and looks at the pattern instead of one result.

Should I try to game the answer by mentioning my business in a lot of places online?

Not the way that phrase usually gets sold. Buying mentions or spamming directories is a fast way to damage the exact signals AI engines use to trust a listing. The lever that actually moved in our research was ordinary Google presence for the question, which means real reviews, an accurate Google Business Profile, and a page per service and town that a customer would recognize as useful.

Once you have your notes from the manual check, the honest next step is looking at the questions where your business did not appear and asking why Google itself does not surface you for those queries. That is the biggest measured lever in the system, and it is the same work that shows up across every engine. If you would rather have the eight-questions-per-engine version done for you, the free AI Visibility Check is the shortcut. Either way, you are answering the same question your customers already are.