9 Steps to Track Brand Mentions on ChatGPT (2026)
To track brand mentions on ChatGPT, ask it the questions your customers ask in a clean chat, log whether your brand shows up and which sites get cited, then rerun the same prompts every month. A spreadsheet and 20 minutes are enough to start, and a tracking tool can take over once you outgrow the sheet.
Below are the nine steps I run for clients, with the exact prompts and what to log.
Key Points
- Test in a Temporary Chat with memory off, since ChatGPT shapes answers around what it knows about you
- Category prompts (“best [category] for [use case]”) decide whether you get recommended, so track those first
- Log four things per prompt: mentioned or not, position, sentiment, and the domains cited
- Wrong answers trace back to pages ChatGPT reads, so the fix is usually on a website, not in the chat
1. Start a Clean Chat
Start every check in a Temporary Chat or with memory turned off, because ChatGPT personalizes answers using what it has saved about you. OpenAI’s memory FAQ confirms that Temporary Chats do not use existing memories or create new ones, which makes them the closest thing to a stranger asking.
If you have custom instructions set, turn those off too. A profile that says “I run a skincare brand” nudges every answer toward your world, and that is not what a shopper sees.
Also, run the check logged out in a second browser if you want a second opinion. The two answers usually match, and when they do not, the logged-out one is the one to trust.
2. Ask the Brand Questions Customers Ask
Ask ChatGPT the questions your customers already ask about you, in their words. Nobody types “leading provider of premium solutions.” They ask if the thing is good, whether it is worth the money, and what else they should look at instead.
My standard brand prompt set:
- What is [brand]?
- Is [brand] legit? What do reviews say?
- [Brand] vs [main competitor]: which is better?
- Where should I buy [your product type]?
Question two earns extra attention, because ChatGPT reads public reviews and quotes them back. If the answer leans on one bad review from 2023, you have just found your first fix.
3. Ask the Category Questions
Ask the category questions next, because “best [category]” prompts decide whether ChatGPT recommends you at all. Shoppers who already know your name will find you. The ones asking for the best option in your category are the ones you win or lose inside the answer.
Word them the way a buyer would: “What are the best [category] for [use case]?”, “Which [category] would you recommend under [price]?”, and “What are the top [category] brands?” Swap in the two or three use cases your best customers actually buy for.
The stakes on those answers keep climbing. Pew found 44% of U.S. adults now use ChatGPT, up from 18% in 2023, and more of them ask it what to buy every month. A brand that never shows up in category answers is invisible to those shoppers.
4. Turn On Search and Open Every Source
Run the same prompts with web search on, because sourced answers show you where ChatGPT gets its information. OpenAI’s search docs explain that it rewrites your question into search queries, then cites the pages it leans on.
Open every citation and write the domain down. That short list is your brand’s information supply chain, the same handful of sites I map when I look at how ChatGPT picks which sites to cite. Most brands I audit are shaped by fewer than ten pages.
OpenAI itself warns that citations can be incomplete or outdated and tells users to open the source and check it. Do that for each one. A page that praises you is worth strengthening, and a page that gets you wrong just became your to-do list.
5. Run the Same Prompts With Search Off
Run each prompt a second time with search turned off, because the no-search answer shows what the model believes from training alone. The searched answer shows what the live web adds on top.
When the two disagree, note which one is wrong. A wrong no-search answer means old information baked into the model, which only time and fresh sources fix. A wrong searched answer usually points straight at the page causing it.
Keep both versions in your log. Over a few months you will see the training-only answer slowly catch up to the searched one, which tells you your fixes are landing.
6. Log Every Mention in a Tracking Sheet
Log each answer in a spreadsheet, one row per prompt and one column per month. Paste in the answer, then score it. Four columns do most of the work: mentioned or not, position in the list, sentiment, and the domains cited.
- Mentioned: did your brand appear at all
- Position: first, third, or buried at the bottom
- Sentiment: recommended, neutral, or warned against
- Sources: the domains ChatGPT cited for that answer
Add a fifth column for competitors named, since a new rival showing up in “best of” answers is an early warning. Screenshots go in a folder with the date in the filename, because you will want the receipts later.
7. Fix Wrong Claims at the Source
Fix what ChatGPT gets wrong by fixing the pages it reads, not by arguing with the chatbot. Wrong prices come from stale product pages. Wrong descriptions usually trace to an old About page, a dead directory listing, or a third-party writeup nobody has touched in years.
Still, some errors are not sourced from anywhere, they are invented. That failure has its own causes and fixes, which I covered in why ChatGPT makes things up about brands. For everything else, there is a page to blame.
Update your own site first, then the third-party pages that keep showing up in your citation list. A current About page, clear product pages, and an FAQ that answers “is this brand legit” do more for your AI answers than any trick. ChatGPT is only as current as the pages it can read about you.
8. Add a Tracking Tool When the Sheet Gets Old
Add a tracking tool once you are running more than 20 or 30 prompts, because manual checks stop scaling around there. Tools run your prompt set daily across ChatGPT, Perplexity, and Google AI Mode, then chart mentions and share of voice over time.
I compared the main options in the best AI visibility tools. Most charge a monthly fee, and the cheaper tiers cover one brand and a few dozen prompts, which is plenty for a single store.
Just know the tool does not replace the manual check. It counts mentions well, but it cannot tell you the page behind a wrong claim the way opening a citation yourself does. Run the sheet for two months first so you know which prompts matter before you pay to automate them.
9. Repeat Monthly and Judge the Trend
Repeat the full prompt set every month, because a single check is a snapshot of a moving target. AI answers to the same question can swap out half their sources within weeks, a churn problem I measured in whether AI citations are stable.
Judge the trend line, not the screenshot. One bad answer in one chat is noise. The same wrong claim three months in a row is a source problem you can trace and fix, and a lost recommendation two months running means a competitor took your spot.
Most brands skip this entirely. Semrush’s 2026 AI Visibility Index found 45% of marketing leaders cannot accurately measure their brand’s visibility in AI answers, and only 9% have tools that track it across platforms. Twenty minutes a month puts you ahead of both groups.
Want Your ChatGPT Brand Audit Done for You?
Tracking brand mentions on ChatGPT comes down to nine repeatable steps: a clean chat, the right prompts, open sources, a simple log, and a monthly rerun. Almost nobody runs that loop, and it is the cheapest way to know what half your future customers are being told about you.
Book a call and I’ll run the full prompt set for your store, hand you the list of pages shaping your AI answers, and flag the fixes worth making first. If it turns into a bigger project, my AI search services start with exactly that work.
