How to check if ChatGPT recommends your business (and get an answer you can trust)

Most business owners have already done the quick version: open ChatGPT, type “best [your service] in [your suburb]”, and see who comes up. The trouble is that AI answers change between identical runs, so one check tells you very little. (We explain why in Why ChatGPT gives a different answer every time.)

Here is a method that gives you an answer you can actually rely on, using nothing but the free versions of the AI tools and a spreadsheet.

Step 1: Write the questions your customers ask

Start with the decision a customer is making, not a keyword. A homeowner choosing a battery installer might ask:

• “Who is the best solar battery installer in Newcastle?”
• “Is a home battery worth it in NSW with the rebate?”
• “Which battery installers near Lake Macquarie have good reviews?”

Write five to ten questions across one or two of your most valuable services. Include the location the way a customer would, and mix “who is best” questions with “should I” and “how much” questions, because AI often recommends providers in answers to those too.

Step 2: Choose the platforms

Your customers do not all use the same assistant. Check at least:

• ChatGPT
• Google AI, meaning AI Overviews and AI Mode inside Google Search
• Gemini, Google’s standalone assistant, which can answer differently from Google Search
• Perplexity

Step 3: Ask each question several times, cleanly

This is the step most people skip, and it matters most.

• Use a fresh chat for every run, so earlier messages do not steer the answer.
• Where you can, use a logged-out or private window, or turn off memory and personalisation. Your own history can make an assistant more likely to mention your business.
• Ask each question at least five times per platform. More is better.

Step 4: Record every answer

Make a spreadsheet with one row per answer:

Example row with fictional businesses.

In the “You” column, use one of three values:

• Recommended: the answer actively puts you forward.
• Mentioned: you are named, but not recommended, for example in a list of “other options”.
• Absent: you are not there at all.

Keep the full answer text somewhere too. You will want it later.

Step 5: Read the pattern, not the single answer

Once you have a few dozen rows, count:

• How often you were recommended, per question and per platform.
• Which competitors were recommended most often, and on which questions.
• Which sources keep being cited when competitors are recommended and you are not.

That is your baseline. A gap on one platform and not others, or one competitor leading every “who is best” question, tells you where to focus.

Common mistakes

• Trusting one answer. It is a sample of one.
• Checking while logged in to your own account. Personalisation can flatter you.
• Counting a mention as a recommendation. It hides the gap that matters.
• Changing the questions every time. You cannot compare results if the questions keep moving.
• Assuming a change you made caused a later improvement. AI answers move for many reasons. Compare before and after periods, measured the same way, and watch what competitors did over the same time.

When to automate it

The manual method works for a one-off check. It becomes hard to keep up when you want to track several services, several platforms and a handful of competitors every week, or when you need to show a client or a business owner what changed after some work.

That is what Ansoar does: it runs every tracked search once a day on each platform you choose, keeps every answer as evidence, separates recommended from mentioned, shows which competitors lead and which sources they are cited on, turns the gaps into prioritised actions, and measures the same searches after you act. Agencies can do this for every client from one account: see Ansoar for agencies.

Ben Rosenbaum is the founder of Ansoar. More about Ben.