How to check whether ChatGPT recommends your business — by hand, in ten minutes

You can check whether ChatGPT recommends your business without any tool. It takes about ten minutes, and it will tell you more than most dashboards do — as long as you do it properly. Here is the method, and the three mistakes that make the result meaningless.

Step 1: write the questions a customer would ask

Do not ask “what do you know about Cedar Park Dental?”. Nobody who does not already know you asks that. Ask what a stranger asks:

• “best dentist in Cedar Park for a family”
• “dentist near Cedar Park TX that takes new patients”
• “emergency dentist open Saturday north Austin”
• “how much does a crown cost in Austin”

Write five to ten of these. Mix general, urgent and price questions. Include your city or neighborhood the way people say it, not the way your address reads.

Step 2: ask each question three times, in fresh chats

This is the step everyone skips. Language models sample their answers. The same question can produce a different list of practices on a different run. If you ask once and see yourself, you might have been lucky; if you ask once and do not, you might have been unlucky.

Open a new chat for each attempt so previous answers do not influence the next. Ask the same question three times. Write down, for every run:

1. Were you named at all?
2. If yes, in what position — first, second, third, later?
3. Who else was named?
4. What did it say about you — address, hours, services — and was it correct?
5. Did it give a link, a phone number, or a way to book?

Step 3: repeat on Perplexity and Google

ChatGPT is one engine. Perplexity answers with citations, which tells you which pages it trusted. Google's AI Overviews appear at the top of a normal Google search and reach people who never open a chat app. Run the same questions on both. On Perplexity, note the sources it cites: those are the pages deciding your visibility in your market.

Step 4: read the table you just made

You now have a small table: five to ten questions, three engines, three runs each. Look for three patterns.

You are absent everywhere. The usual causes are technical: your robots.txt blocks AI crawlers, your content is only rendered by JavaScript, or your business is not on the directories the engines cite. Check the robots.txt guide first.

You appear sometimes. This is the normal state for most local businesses, and it is where measuring matters. A mention in one of three runs is a 33% mention rate — but with three samples the honest range is wide. This is why Anserra reports intervals instead of single numbers.

You appear, but with wrong data. Old hours, an old address, a service you no longer offer. Note where the engine got it from (Perplexity will tell you). Fixing the source fixes the answer.

The three mistakes

Asking about yourself by name. A brand question tells you the model knows you exist. It does not tell you whether the model recommends you to someone who does not.

Asking once. One answer is one sample. Treat it as a coin flip until you have several.

Asking from your own account. If you are logged in and have discussed your business with ChatGPT before, memory features can bias the answer. Use a fresh chat, or a logged-out session where possible.

What a tool adds

The manual check tells you where you stand today. It does not tell you next week, does not test whether an AI agent can actually book with you, and becomes a chore at 25 questions across three engines. That is what the free scan does in a few minutes — and the paid plans repeat it weekly, with the agent test and the fix files.

But run the manual check first. It takes ten minutes, and it makes every number you see afterwards mean something.