How to Check Whether ChatGPT Mentions Your Brand (A Simple Tracking Method)

To check whether ChatGPT mentions your brand, write a fixed list of the questions your buyers actually ask, run each one in a fresh, logged-out session on the same day every month, and record four things in a spreadsheet: whether you were named, where you appeared in the answer, the exact words used about you, and which sources were cited. Do the same in Perplexity, Gemini and Google. One check tells you almost nothing. The same check, repeated the same way for a few months, tells you a lot.

The rest of this post is the method, step by step, so you can start this week without buying a tool.

Why a single test is misleading

AI answers are not fixed. The same question can produce a different answer an hour later, from a different account, or in a different city. ChatGPT may also remember things about you if you are logged in, and it will happily mention your brand back to you because you talked about it last week.

So a founder types “best organic dog treats” into ChatGPT, sees their brand, and relaxes. Or doesn’t see it and panics. Neither reaction is earned. What you want is a trend measured the same way each time, so the only thing that changes between checks is the world, not your method.

Step one: write a fixed prompt list

Write between ten and twenty prompts and freeze them. Do not edit them later, or your history stops being comparable. If you want to test new prompts, add them as new rows and leave the old ones alone.

Cover three kinds of question:

• Category questions. “What are the best [product type] brands?” or “Which [product type] is worth buying?” These are where you most want to be named.
• Problem questions. “How do I fix [problem your product solves]?” The answer may recommend products, or it may not. Either way you learn how the model frames the problem.
• Comparison and brand questions. “[Your brand] vs [competitor]” and “Is [your brand] any good?” These show what the model believes about you when it is asked directly.

Write them the way a customer would type them, not the way your marketing team would. Short, a little vague, sometimes with a budget or a use in mind. Pull wording from your reviews, support emails and search terms if you can.

Step two: control the session

This is the part most people skip, and it is why their checks disagree with each other.

• Use a private or incognito browser window.
• Stay logged out where the tool allows it. If you must log in, use a clean account that has never discussed your brand, with memory and personalization turned off.
• Start a new chat for every prompt. Never ask two prompts in the same thread, because the second answer is shaped by the first.
• Note your location. Answers can change by country and sometimes by city. If you sell nationally, check from the same place each time.
• Use the default model setting the tool offers to a normal user, and write down which one it was.

Step three: pick a day and keep it

Choose a fixed day, such as the first Monday of each month, and run the whole list that day. Monthly is enough for most brands. AI answers move, but your efforts to change them move slowly, and checking daily mostly measures noise.

If you have the time, run each prompt twice in separate fresh sessions on the same day. If the two answers disagree, record both. That disagreement is useful information about how settled the model is.

Step four: log it in a simple spreadsheet

One row per prompt, per tool, per check. Use these columns:

• Date
• Tool (ChatGPT, Perplexity, Gemini, Google AI Overview, Google AI Mode)
• Model or mode (whatever the tool showed)
• Prompt (copied exactly from your frozen list)
• Named (yes or no)
• Position (first brand named, second, third, later, or not named)
• Wording (the exact sentence about you, pasted in)
• Accurate (yes, no, partly, with a note on what was wrong)
• Competitors named (in the order they appeared)
• Sources cited (every link or source the answer showed)
• Screenshot (file name or link)
• Notes

The Wording column matters more than people expect. Being named as “a budget option” or “a brand with mixed reviews” is a different result from being named as “a favourite for sensitive stomachs.” Over time you will see whether the description is drifting toward what you want buyers to hear.

The Sources column is where the action is. If the answers keep citing the same review site, the same Reddit thread or the same comparison article, those are the pages shaping what the model says. That tells you where to spend effort next. We cover why that matters in our post on why consistent mentions across the web matter.

Step five: take screenshots every time

Screenshot every answer, full length, with the date visible if you can. Save them in a folder per month and name the files plainly, such as “chatgpt-best-dog-treats-october”. Screenshots settle arguments later. They also catch things a spreadsheet misses, like a product card, a map, or a shopping carousel that showed your competitor’s photo.

Step six: check beyond ChatGPT

ChatGPT gets the attention, but your buyers use other tools too. Run the same frozen list in:

• Perplexity. It shows its sources clearly, which makes it the easiest place to see which pages are feeding answers.
• Gemini. Answers can differ from Google Search, so treat it as its own tool.
• Google Search. Record whether an AI Overview appeared at all, whether you were named in it, and which links it showed. Note your normal organic position too, since the two do not always match. Our guide on how Google AI Overviews choose sources explains why.
• Google AI Mode, where it is available to you.

If you sell to businesses, add Claude and Microsoft Copilot as well.

How to read the results

After three or four monthly checks, look for patterns rather than single wins:

• Share of prompts where you are named, by tool. Is it rising, flat or falling?
• Position. Moving from “named last” to “named first” on category prompts is real progress.
• Accuracy. Wrong prices, discontinued products or an old company description usually trace back to a stale page somewhere on the web. Find it and fix it.
• Recurring sources. The pages cited again and again are your shortlist for outreach, corrections and new content.
• Competitors that keep appearing. Look at where they are mentioned that you are not.

Be honest with yourself about what this measures. It is a sample, not a census. Real users ask longer, messier questions, and their answers are personalized. A rising trend on a fixed list is a good sign. It is not a promise that every buyer sees you.

When to hand it off

This method takes a person a couple of hours a month once it is set up. Many owners do it themselves for a while, which is a good way to learn what the tools think of you. If you later hire help, ask the agency to show you their tracking method in this level of detail. A vague “visibility score” with no prompt list behind it is a warning sign, and our list of questions to ask before hiring an AI SEO agency goes into that. You can also see how agencies compare in our ranking of AI SEO agencies.

The short version

• Freeze a list of real buyer prompts.
• Run them in fresh, logged-out sessions, one chat per prompt.
• Same day each month, same location.
• Log named, position, wording, accuracy, competitors and sources.
• Screenshot everything.
• Check Perplexity, Gemini and Google too.
• Judge the trend, not the single answer.