How to Check Whether ChatGPT Recommends Your Business

Most business owners have never once asked an AI assistant what it says about their business. They have opinions about their ChatGPT visibility. They have a vague worry about it. But they have never sat down, typed the question a customer would type, and read the answer back.

That is a strange gap, because it is the cheapest research you will ever do. It costs nothing, takes twenty minutes, and it is the only way to replace a worry with a fact. This is how to run a ChatGPT visibility check properly, and just as importantly, how to read what comes back without fooling yourself in either direction.

Why One ChatGPT Visibility Check Tells You Nothing

Here is the part most people get wrong before they start. They ask once, see their name, and conclude they are fine. Or they ask once, do not see their name, and conclude they are invisible. Both readings of your ChatGPT visibility are worth about the same: very little.

AI assistants do not return a fixed ranking the way a search results page does. The same question can produce different answers in two consecutive sessions. Answers vary by how the question is phrased, by what the assistant knows about the person asking, by whether it chose to search the web that time, and by which sources it happened to pull. Two customers in the same town can ask the same thing and get different businesses back.

So a single check on your ChatGPT visibility is a snapshot, not a scoreboard. That is not a reason to skip it. It is a reason to do it properly: several prompts, recorded, repeated over time. One run tells you almost nothing. Twelve runs over three months tell you something real.

This matters for your sanity as much as your marketing. If you treat one bad answer as proof you are failing, you will make panicked decisions. If you treat one good answer as proof you have won, you will stop doing the work that earned it.

The Five Prompts Worth Running

The instinct is to type your business name. Resist it. Asking an assistant about a business you have named is a memory test, and it usually passes. Customers do not search that way unless they already know you, and if they already know you, you have not learned anything.

Ask the way somebody who has never heard of you would ask. Five prompts, each in a brand new chat:

• The straight recommendation. “Can you recommend a [what you do] in [your town or area]?” This is the closest thing to the money question.
• The problem, not the service. “I need help with [the problem your customer actually has]. Who should I talk to in [area]?” Customers describe symptoms, not service categories.
• The comparison. “What are the best options for [what you do] near [area], and how do they differ?” This shows you who the assistant considers your competitive set, which is often not who you think.
• The qualifier. “I am looking for a [what you do] who is good with [your particular strength, your niche, your type of client].” This tests whether your specialism has landed anywhere the machines can read it.
• The direct check, last. “What can you tell me about [your business name] in [area]?” Run this one last, in its own fresh chat, and treat it as a fact-check rather than a visibility test. You are looking for whether what it says is correct, not whether it says anything.

Run all five in ChatGPT first, since it is the assistant most of your customers will have. It reported more than 800 million weekly users in 2026, which makes it the default for a very large number of people. Then repeat the set in whichever others matter to you. Perplexity and Google’s AI answers are the usual next two.

How to Record What You Find

This is the step that separates a useful habit from a nervous scroll, and almost nobody does it.

Screenshots are not enough. You want the words, because in three months you will want to compare the words. Open a simple document or spreadsheet and, for every single run, capture five things:

• The date. Answers drift. Undated evidence is not evidence.
• The engine. ChatGPT, Perplexity, Google, Copilot. Never write “AI said” without naming which one.
• The exact prompt. Word for word, because a small rewording changes the answer.
• The full answer, pasted as text. Not a summary of it. Not your impression of it. The verbatim answer.
• Who was named, and in what order. Including the competitors. Especially the competitors.

Twenty minutes, once. Then you have a baseline, and every future ChatGPT visibility check has something to be measured against. Without a baseline you are permanently guessing, and guessing feels like knowing, which is the dangerous part.

Save the competitor names too. Over several months, one or two will keep appearing. Those are the businesses the machines currently trust in your category, and looking at what they publish is more instructive than any amount of theory.

Reading the Results Honestly

When you read back what you collected, sort every result into one of three buckets. Three, not two. The third is the one people skip, and skipping it is how you end up with a confident, wrong picture.

• Mentioned. You were named, in this run, for this prompt. Note where in the answer, and what it said about you.
• Not mentioned on this check. You were not named in this run. That is the honest phrasing. It is not the same as “we are invisible,” and writing it the careful way keeps you from over-reacting.
• Could not verify. The answer was vague, refused, hedged, or gave you something you cannot confirm either way. This bucket is real and it is common. Forcing it into one of the other two is where people start lying to themselves.

Then look at accuracy separately from presence. Being mentioned with the wrong opening hours, an old address or a service you stopped offering two years ago is not a win. It is a different problem with a different fix, and it is usually a faster one.

Read what it says about the businesses it did name, as well. Notice the reasons it gives. “Known for”, “frequently recommended for”, “specialises in”. Those phrases tell you what kind of evidence the assistant found convincing, and that is a straight instruction for what to build.

What to Fix When You Are Not Mentioned

Nothing that improves ChatGPT visibility is exotic. The fixes are unglamorous, which is exactly why they are available to you.

Start with your facts. Name, address, phone, hours, services and area covered, identical everywhere they appear. Assistants are pattern matchers pulling from many sources at once, and contradictions make you a risky thing to recommend. Your Google Business Profile carries disproportionate weight here and is the first place to look.

Then your evidence. Detailed, recent, specific reviews teach the machines what you are actually good at, in a way a star rating never can. I wrote about why reviews now function as training data and how to earn the kind that carry information.

Then your answers. Take the five prompts you just ran and notice which ones your website does not properly answer. Write those answers. Question as the heading, straight answer in the first paragraph, detail underneath. Three genuine answers beat twelve thin posts, and this is the whole idea behind answer engine optimisation.

If you want the full sequence rather than the summary, the AI visibility checklist lays out what to work through, and what AI visibility actually means covers the groundwork if the whole area is new to you.

How Often to Repeat Your ChatGPT Visibility Check

Monthly is plenty for most businesses. Same five prompts, same engines, same document, first working day of the month. It takes twenty minutes and it turns a vague anxiety into a line on a chart.

Check more often only if you have just changed something significant, a new location, a rebrand, a merged listing, or a burst of new reviews. Then a fortnightly check for a couple of months will show you whether the change landed.

Do not check daily. You will be reading noise as signal, and the variance between runs will drive you slightly mad. The whole value of tracking ChatGPT visibility is in the comparison over months, not in any single answer.

The Takeaway

You cannot improve what you have never looked at. Five prompts, five fresh chats, one dated document, repeated monthly. That is the entire ChatGPT visibility practice, and it puts you ahead of almost every business in your category, most of whom are still guessing.

Be careful about what you conclude from any single run, in either direction. Variance is normal, personalisation is real, and the honest unit of measurement is the trend across months, not the answer in front of you today.

If you would rather have this run properly and tracked for you, you can book a dashboard demo with The GEO Agency and see the multi-engine version, with saved answers and month-on-month comparison. If it is your own thinking you want a hand with rather than your marketing, that is what my one-to-one work is for.

Ayse Durmush

AI Visibility Expert & Global Business Consultant

Ayse has spent three decades transforming strategy, lead generation and operations for businesses of every size, and has consulted for high profile global brands. She writes about AI visibility, mindset and building a business that can survive its founder.

More about AyseWork with Ayse one to oneAll articles

Keep reading

BusinessWhat to Review Before September StartsBusinessThe Trust Signals AI Looks For Before Recommending YouBusinessWhy AI Assistants Disagree About Your Business