Does ChatGPT Name Your Local Business? How to Check

Ask ChatGPT for “a house painter in Denver” and the answer names a few businesses, often with a star rating and a review count. If yours isn’t one of them, that customer is choosing from a list you’re not on, and you may never see it happen.

This post shows how to check by hand whether ChatGPT, Gemini and Google AI Mode name your local business, what to write down, and why one try tells you little. If you would rather not do it by hand, the free local check runs one question for you with no account.

Why it is worth checking

We found no figure for how many customers ask an assistant before they call that is good enough to quote, so this post quotes none. What you can learn is whether the assistants name you when someone asks the question your customers would ask.

Check it by hand

1. Write the question the way a customer would, with your town in it and your business name left out. “Who are the best house painters in Denver, Colorado?” tests whether you get named. “Is [your business] any good?” only tests whether the assistant has heard of you. Type the town yourself instead of relying on “near me”, so you know which place the answer is about.
2. Keep to one service per question. A business that paints houses and fences has two questions, not one.
3. Open a new chat for each try, so the last answer can’t steer the next one. If the assistant has a temporary or incognito mode, use it.
4. Read the whole answer from top to bottom. Are you named? Which other businesses are? Does it show a rating and a review count for each? Which websites does it link to?
5. Ask on more than one assistant: ChatGPT, Gemini (Google’s assistant) and, if you have it, Google AI Mode (the AI mode in Google Search). They answer differently, so a result on one doesn’t tell you what the others will say.
6. Ask each question at least three times, on different days if you can, and change the wording a little, for example “house painters” and then “exterior painting”.

What to write down

A plain table is enough. One row per try. The rows below are an invented example for a house-painting business in Denver, not real data, and the business does not exist.

The first two rows are the same question on the same day, with different results. At the bottom, count: named you in 1 of 3. This example mixes two assistants and two wordings, so once you have more rows, count each assistant and each question on its own. A record like this is a count of real answers, not a feeling. By hand, the full set (a few questions, three assistants, three tries each) means dozens of asks, which is why the repeated version below exists.

Why one try tells you little

Two askings of one question can return two different lists. SparkToro and Gumshoe.ai’s January 2026 study put the chance of two responses giving the identical list of brands below 1 in 100, for ChatGPT and for Google. It covered 12 prompts, two of them tied to a place (Volvo dealerships in Los Angeles and cancer care hospitals on the West Coast). The details are in why we ask the same question 8 times a week.

Our own local samples looked different, with limits, and they measured something looser. For our dated samples of US law firms, dental practices and accounting firms, we asked each question twice on the same day. We counted what share of the shorter answer’s business listings, the cards ChatGPT shows for each business, matched by business name, also appeared in ChatGPT’s other answer. The median share was 0.94, 0.80 and 0.82, counting only the questions where both of ChatGPT’s answers showed listings (24, 37 and 38 of the 40 asked in each sample), where 1 means all of them. In 12 of 24, 9 of 37 and 11 of 38 pairs, every listing in the shorter answer also appeared in the other. The same set of listings, in any order, came up in only 0, 2 and 2 pairs, and 4 of 24, 2 of 37 and 2 of 38 pairs shared fewer than half their listings. In plain terms, ChatGPT’s local answers repeated much of the same list, but they were rarely identical. SparkToro asked how often two responses gave the exact same list, so the two numbers don’t contradict each other. Ours is one day and two asks. It says nothing about next week or about Gemini and Google AI Mode, which we did not read for business names. The samples are law firms, dental practices and accounting firms.

So a single answer that names you may be luck, and a single miss may be bad luck. Keep a count: out of many answers, how many named you.

Reading what you found

A business that keeps turning up is who you’re being compared with, whether or not you knew it. Directories, review sites and maps entries that show up among the sites an answer links to are different: they are places an answer drew on, not rivals, so count them separately. And being named is not the same as being recommended. An answer can name you and lead with someone else.

This post stops at measuring, and so does the local version of Bevizio: it shows what the assistants said and does not tell you what to change. If it keeps being a competitor and not you, why AI engines name your competitor instead of you sets out what is known about it.

What this cannot tell you

A hand check can’t tell you how many customers use AI, or whether being named brings calls. Your own check can also differ from what a customer sees, because of account, location, model version and time of day. Bevizio isn’t affiliated with OpenAI, Google or Anthropic.

The faster, repeatable version

Bevizio is a tool that does this counting for you. The free local check does the first steps. Give it your business name, your kind of business, your town and country, and it asks ChatGPT, Gemini and Google AI Mode one question, without your business name in it. It shows who each answer named, with the rating and review count where the assistant gave them. That is one answer from each assistant, a dated snapshot and not a verdict on your business.

The repeated version comes with the 14-day trial, which needs a card. Each question, one service in one town, is asked 8 times a week: 3 times on ChatGPT, twice on Gemini, twice on Google AI Mode and once on Claude, Anthropic’s assistant. Bevizio reports a count such as “named in 2 of 3 answers”, meaning 2 of the 3 answers so far named you, and keeps small numbers as raw counts. A painter who does interior and exterior work in two towns has four questions. Plans are $29 a month for 20 questions or $79 for 75, and are compared on pricing. How the answers are collected is on how we measure.

For the wider idea, what is AI visibility, and how is it measured? defines the term. If you sell software and not a service in a town, does ChatGPT recommend your product? is the guide for you.