Check if assistants mention your business. A method anyone can run by hand.

Write five to ten questions a customer would type, without your business name. For example: "Who repairs garage doors in Boise, Idaho?" or "Which bakery in Red Hook is open on Sunday?"

Put the city and state in each question. OpenAI's help page says ChatGPT guesses location from the IP address, and suggests adding a city or postal code when local results show the wrong area.

Add one question with your name in it, such as "What are the hours of [business] in [city]?" It shows whether the assistants have your facts right.

• Start a new chat for every question. Earlier messages change the answer.
• Sign out, or turn memory off. OpenAI says ChatGPT may use saved memories when it rewrites a search query.
• Make sure web search is on.
• Ask each question at least five times in each assistant, spread over a few days.
• Write down three things each time: was the business named, were its facts right, and which sources were cited.

Because the lists change. SparkToro and Gumshoe had 600 volunteers run 12 prompts 2,961 times across ChatGPT, Claude and Google's AI in late 2025. The chance of getting the same list twice was under 1 in 100, and the same list in the same order about 1 in 1,000.

The share of answers that named a brand was much steadier. The authors call that share, measured across dozens to hundreds of prompts run several times, a reasonable metric. Gumshoe sells an AI tracking tool, and the study was not peer reviewed.

Five runs give a rough picture: never, sometimes or usually.

• Named in most runs: the business is in the set the assistant draws from for that question.
• Never named: check that crawlers can read your page and that it is indexed. See what a page needs to be quoted.
• Named with wrong facts: open the cited source and fix the fact there. The assistant repeats what the source says.
• First or third in the list: it means little. The order changes from run to run.

For Google, Search Console has a report on generative AI features, described in Google's AI guide.

Server logs show fetches by ChatGPT-User and Claude-User. OpenAI's crawler overview and Anthropic's crawler page say these agents open a page when a person's question needs it.

Website analytics can show visits that arrive from an assistant. That count is low by nature. Pew Research Center tracked 900 US adults in March 2025 and found people clicked a link inside a Google AI summary in about 1% of the visits that showed one. People read the answer and call without clicking.

Answers change with the person's location, account, wording and the model version. Your runs are a sample, not the whole picture.

Paid trackers have limits too. Many query the assistants through an API, which may not match what people see in the apps. The SparkToro authors list this as an open question. Google's AI guide says: "No third-party tool has access to our internal ranking or AI systems."

A check shows whether a business appears. It does not show why.

The Pro plan sends a monthly report on the OnMap page itself: what AI assistants read on it, the visits they sent, and calls and texts. It does not replace the hand check above. See pricing.