How to check if ChatGPT recommends your business

To check whether ChatGPT recommends your business, ask it the questions a customer would ask, without your business name in them, and repeat each question three times in fresh chats. Log the date, the assistant and which businesses it named. The result is a share, such as “named in 2 of 3 answers”, not a yes or no.

Start with questions that do not contain your name

The most useful test questions are the ones a stranger would type. A person looking for a plumber does not know your company name yet, so a question that includes it tells you nothing about whether you get found. It only shows what the assistant has picked up about that name.

Write five to eight questions for your own trade and town. Here are some examples to adapt:

• “Which hair salon in Austin would you recommend for balayage?”
• “I need an HVAC company in Denver that can replace a furnace this week. Who should I call?”
• “Who is a reliable plumber in Leeds for a boiler repair?”
• “What is a good family dentist near the centre of Cork?”
• “Which independent bakery in Portland has good reviews?”

Vary the wording on purpose. Customers ask short questions, long questions and questions with a detail attached, like a price range, an emergency or a neighborhood. Assistants can answer each version differently, so a few variations per service give you a fairer picture than one perfect prompt.

Use a fresh chat every time

Open a new chat for every attempt, or use a temporary chat. If you reuse a long conversation, earlier messages can shape the next answer. If your account has memory turned on, ChatGPT may also carry details about you into the reply, including things you told it about your own business.

Also think about where you are asking from. Some assistants take your approximate location into account for “near me” style questions. If you ask from your shop in Austin, you may see something different from what a customer in another part of the city sees. For this reason, put the town or neighborhood in the question itself instead of relying on location.

Repeat each question three times

Run every question at least three times, in separate chats, ideally on different days. Answers are generated each time, so the list of names can change from one attempt to the next even when nothing about your business has changed.

That is why one attempt cannot support a conclusion in either direction. If you were named once, you cannot say you are reliably recommended. If you were not named once, you cannot say the assistant ignores you. Three attempts give you a rough rate, and more attempts over several weeks give you a clearer one.

Keep a simple log

Without a record, the test is just an impression. A spreadsheet is enough. Add one row per attempt:

Date	Assistant	Question	Your business named	Number of businesses named
Oct 27	ChatGPT	Which hair salon in Austin would you recommend for balayage?	no	4
Oct 27	ChatGPT	Which hair salon in Austin would you recommend for balayage?	yes	3

The date and the assistant matter because both change. Providers update their models, and the sources an assistant draws on shift over time. A result without a date and a tool name cannot be compared with a later one.

When you summarize, write it as a count: “named in 1 of 3 answers on Oct 27 in ChatGPT”. If it never appeared, write exactly that: “in none of the 3 answers on Oct 27 was it among the businesses named”. Avoid turning that into “ChatGPT does not recommend us”. You can confirm what showed up. You cannot confirm what is absent everywhere.

Look at the sources, not only the names

When a ChatGPT answer includes links or cited sources, open them. They show which pages the assistant relied on for your trade and town. For a hair salon in Austin, that might be a local directory, a review site or a city guide. For a plumber in Leeds, it might be a trade register or a community forum thread.

This is often more useful than the list of names. If the same few sites keep appearing, you know where your information needs to be accurate and current. Check that your business is listed there, that the name, address and phone number match your website, and that the description says plainly what you do and where.

OpenAI documents a crawler called OAI-SearchBot, which it says controls whether a site appears in ChatGPT search answers. If your site blocks it in robots.txt, it can affect whether your own pages show up as sources. OpenAI explains the details in its crawler documentation.

The app and the API are not the same thing

What you see in the ChatGPT app is not necessarily what you get from the same question sent through the API, the programming interface developers use. The app can involve memory, saved instructions, your location and a web search step that you or the product turned on. A plain API call may have none of those, or may have them configured differently.

Neither is wrong. They are different measurements. A customer typing into the app sees the app experience, so the manual test above is the closest match. Any tool that uses the API, ours included, is reporting a related but separate thing, and it should say so. Whenever you write down a result, note where it came from.

How Zmienka does this

The manual method works, but it takes time and the results are hard to compare month to month. Zmienka automates the same idea. You enter your website address, we work out your trade and town from the site, and we ask an AI assistant the questions a customer would ask, without your business name. Each question is asked three times because answers vary.

Right now the assistant we test is Google Gemini with web search. ChatGPT and Perplexity are coming soon. The result reads like “named in 2 of 3 answers”, with the date and the assistant, and the report shows which businesses were named and which pages the answers drew on. Zmienka does not promise positions. It measures where you stand before changes and again afterward.

You can start with the free quick test on the Zmienka home page. If you want to understand how checks like this fit into a longer view, see AI visibility monitoring. For background on the topic itself, read our short guide to what GEO means for local businesses, and if the results are not what you hoped, why assistants mention competitors goes through the common reasons.

What to do with the results

Treat the first round as a baseline. Save the log, make the sensible changes you can control, such as correcting listings, clarifying your website and checking robots.txt, and run the same questions again after a few weeks. The comparison between the two rounds is the useful part. Nothing you do guarantees a mention, but a log lets you see whether things moved.

Frequently asked questions

How do I find out if ChatGPT recommends my business?

Ask it the question a customer would ask, such as “Which hair salon in Austin would you recommend?”, without your business name. Use a fresh chat, repeat the question at least three times and log how often your business is named.

Should I put my business name in the question?

Not for the main test. A customer who does not know you cannot type your name, so the useful question is the one without it. A question with your name only shows what the assistant knows about that name.

Why does ChatGPT give different answers to the same question?

Each answer is generated fresh, and it can also depend on location, chat history, memory settings and whether web search was used. That is why a single answer cannot tell you much.

Is the ChatGPT app the same as what an API test shows?

Not necessarily. The app can use settings such as memory and web search that a plain API call does not, so results can differ. Always note which one a result came from.