How can I check if AI knows about my business?

What are you actually checking?

“Does AI know my business?” is really three separate questions, and each one needs different prompts.

1. Recognition: asked about your business by name, does the assistant find the right business, not a similarly named one somewhere else?
2. Accuracy: are your hours, address, phone number, services and prices correct?
3. Discovery: when someone asks for your kind of business in your area without naming you, are you mentioned at all?

Recognition and accuracy are easier to verify and easier to fix. Discovery is the hardest to influence and the hardest to measure, for the reasons covered below.

Which prompts should you try?

Use plain questions a real customer might type. Replace the bracketed parts with your own details.

Recognition and accuracy prompts

• “What can you tell me about [business name] in [city]?”
• “What are the opening hours and phone number for [business name] on [street]?”
• “What services does [business name] offer, and which areas does it serve?”
• “Is [business name] in [city] still open?”
• “Which sources did you use for that answer?”

Discovery prompts

• “Who are well-reviewed [service] businesses in [neighborhood or town]?”
• “I need a [service] near [landmark or zip code] this week. Who should I call?”
• “Which [business type] in [city] offer [a specific service you provide]?”

Make sure the assistant actually searched the web. Anthropic says you can include “Search the web” in a prompt to make sure Claude uses web search ^[2]. OpenAI says ChatGPT may search automatically when a question benefits from current information, and that you can also start a search manually from the tools menu ^[1]. An answer written without a search may reflect only older training data.

How do you run a fairer check?

• Start a new conversation for each test, so earlier messages don’t steer the answer.
• Try once signed in and once signed out or in a private browser window, and note any difference.
• Name the city or zip code in the prompt. OpenAI suggests including your city, neighborhood or postal code when local results show the wrong area ^[1].
• Turn off any VPN, which can change the location ChatGPT infers from your connection ^[1].
• Ask each question a few different ways. Google’s AI Overviews help page itself suggests asking multiple versions of a question ^[3].
• Open every cited source. OpenAI advises opening a cited source to check that it supports the answer ^[1].
• Repeat the same prompts on different days and keep a simple log.

Why isn’t one AI answer a reliable measurement?

An AI answer is generated fresh each time, from inputs that change. Several documented factors can make the same question produce a different result.

The same question can get a different answer

Language models are not fully predictable. Anthropic’s documentation says that even with temperature set to 0, identical inputs may produce different outputs across API calls ^[4]. Google says AI Mode and AI Overviews may use different models and techniques, so the responses and links they show will vary ^[5].

Location changes the search

ChatGPT may use an approximate location from your IP address and rewrite “near me” questions into searches for a specific city ^[1]. Claude may also use a location inferred from your IP address for local requests ^[2]. A check run from your shop can look different from what a customer across town sees. How AI assistants find local businesses explains how location feeds into results.

Personal settings shape results

If memory is turned on, ChatGPT may use saved memories when it rewrites a search, and saved location information may influence results ^[1]. Microsoft says that with personalization enabled, Copilot remembers key details from conversations and uses them in future ones ^[6]. Your own account, full of past chats about your business, may not be a neutral test.

Products, modes and models differ

Each assistant, and each mode within it, is a different system. AI Overviews appear only when Google’s systems determine they add something to classic Search, and often don’t appear at all ^[5]. Some ChatGPT search features, such as restaurant reservation search, depend on the plan and model in use ^[1].

Answers can simply be wrong

OpenAI says search results and citations can be incomplete, outdated or incorrect ^[1]. Google’s help page says AI Overviews can and will make mistakes ^[3]. A wrong answer may come from a wrong source, or from a mistake in how a correct source was summarized.

Are there better signals than spot checks?

Some platforms now report AI visibility directly, which covers far more answers than a handful of chats. Google says the Generative AI performance report in Search Console shows how content performs in generative AI features on Google Search and Discover ^[7]. Google also says visits from AI Overviews and AI Mode are counted in Search Console’s Performance report, under the “Web” search type ^[5].

Bing Webmaster Tools has an AI Performance report covering Microsoft Copilot, AI-generated summaries in Bing and some partner integrations ^[8]. The Bing report shows citation counts and “grounding queries,” the phrases the AI used to retrieve cited content, and Bing notes that the grounding query data is a sample ^[8].

Be careful with third-party tools that score your “AI visibility.” Google says no third-party tool has access to its internal ranking or AI systems ^[7]. Such tools can still save time by running many prompts at once, but their numbers come from sampled answers and share the limits described above.

What should you do when an answer is wrong?

Look at the cited sources first. If an assistant cites an old directory listing or an outdated page on your own site, fix that source, because the assistant is repeating what it found. Correcting an assistant inside a chat may change that conversation, but we know of no platform documentation saying it updates answers for other people.

If no sources mention your business at all, the problem is usually visibility rather than accuracy. Why ChatGPT can’t find my business walks through the common causes, and how to make your business visible to AI gives an ordered checklist.

A simple log helps you see patterns instead of reacting to one answer. Fill in one row per prompt, per assistant, per session: