← All articles

How to check whether ChatGPT recommends your business

A simple, honest way to see how AI assistants describe and recommend your company today — no tools required, in about ten minutes.

Before you invest in being recommended by AI, it is worth seeing where you stand today. You can do a rough version yourself in about ten minutes. It won’t be a full measurement, but it will tell you whether you have a problem worth solving.

Step 1: write down the questions a buyer would actually ask

Not “tell me about [your company]” — that is a vanity search. Write the questions a potential customer would ask before they know you exist. For example:

  • “Who are the best [what you do] in [your city or region]?”
  • “Which company should I hire to [the outcome you deliver]?”
  • “Recommend a [your category] for [type of client].”

Aim for five or six. These are the moments where a recommendation actually happens.

Step 2: ask each assistant, fresh

Open ChatGPT, Gemini, Perplexity, Copilot and Claude. Use a fresh chat with no memory of you, so the answer reflects what the model genuinely associates with your category — not your own history with the tool. Ask each of your questions and read the answer carefully.

Step 3: record three things

For every answer, note:

  1. Were you mentioned at all?
  2. If so, were you described accurately — right focus, right location, right strengths?
  3. Who was named instead of, or above, you?

Keep it in a simple table. The competitors that keep appearing are the ones the models currently trust more than you. That is useful to know.

Step 4: read it honestly

Three outcomes are common:

  • You are absent. The conversation is happening without you. This is the most common result for good businesses that never worked on it — and the most fixable.
  • You appear but described wrongly. The model knows you exist but has the wrong facts. Often the fastest thing to improve.
  • You appear, described well. Good. Now the job is to hold that position as the models change.

The honest limits of the DIY version

Doing this once, by hand, is a snapshot — not a measurement. Answers vary between sessions, regions and phrasings, so a single run can mislead. Real measurement means asking a fixed set of questions repeatedly, across engines and over time, and comparing before and after a change. That is what we do for clients, and we publish no number we can’t show you.

Want the rigorous version, or help fixing what you found? Start a conversation — or try our guided self-check to see how ready your presence is.


Tell us where you want to lead. We'll tell you, plainly, how we'd get you there.

Talk to us