How to Track Brand Mentions in ChatGPT
Amit Malvi
Founder, SearchAxe. Former Senior SEO Manager, Invideo AI
Published Updated 10 min read
Quick answer
Write a fixed set of prompts the way buyers ask them, run each one several times in an unpersonalized Temporary Chat or logged out, and note whether ChatGPT searched the web. Log whether you were named, recommended and cited, which competitors appeared and which URLs were linked. That is how to track brand mentions in ChatGPT reliably. Report the mention rate monthly and compare it with utm_source=chatgpt.com sessions in GA4.
On this page
- Why are brand mentions in ChatGPT hard to track?
- How should you write prompts for ChatGPT tracking?
- What is the difference between ChatGPT search answers and model answers?
- How to track brand mentions in ChatGPT, step by step
- What should a ChatGPT tracking sheet look like?
- How do you connect ChatGPT mentions to traffic and signups?
- Which tools automate ChatGPT mention tracking?
- How often should you check ChatGPT for brand mentions?
Key takeaways
- ChatGPT rarely returns the same brand list twice, so track the share of runs that name you rather than your position in one answer.
- Memory, location and personalization shape ChatGPT's answers, so run tracking prompts in an unpersonalized Temporary Chat or logged out.
- Log search answers and model answers separately, because only search answers cite URLs and they respond to different work.
- ChatGPT adds utm_source=chatgpt.com to referral links, so the clicks it sends can be matched against the mentions you log.
- Sponsored units are labeled and separate from answers, so exclude them from mention counts.
Tracking brand mentions in ChatGPT means sampling its answers to your buyers' questions on a schedule and turning them into rates. If you're a founder or growth lead at an AI startup and a prospect just told you ChatGPT recommended a rival, this is the method for finding out how often that happens and whether it is changing. ChatGPT is one of seven engines worth watching, alongside Perplexity, Claude, Gemini, Google AI Overviews, Google AI Mode and Copilot; this guide goes deep on ChatGPT alone, and our cross-engine tracking guide covers the rest.
Why are brand mentions in ChatGPT hard to track?
Three things make a single check misleading: answers vary between runs, answers are personalized to the account asking, and ChatGPT sometimes searches the web and sometimes answers from what it learned in training. A tracking method has to control all three before the numbers mean anything.
Personalization is the second trap. OpenAI's help page on searching the web with ChatGPT says that when memory is on, ChatGPT may use saved memories when it rewrites a search query, and that it uses an approximate location from your IP address. A founder who has spent months asking ChatGPT about their own product is the worst possible tester. And since February 2026 OpenAI has been testing ads in ChatGPT on the Free and Go tiers; OpenAI says ads are labeled as sponsored, kept visually separate from the answer and do not influence it, so they should never be counted as mentions.
How should you write prompts for ChatGPT tracking?
Write prompts as a buyer would type them into a chat, with their context, not as keywords. In SparkToro and Gumshoe's research, as Search Engine Journal reported, 142 people asked to write a prompt for the same need produced almost no two alike, yet the engines still drew on a fairly consistent set of brands. So you do not need to guess the exact wording. Add the context a real buyer would (team, budget, stack and goal), keep each prompt fixed once written, and give it an ID.
| Prompt type | Template | Example |
|---|---|---|
| Shortlist with context | Best [category] for [buyer] who needs [outcome] | "What's the best AI video editor for a two-person marketing team that posts daily short clips" |
| Alternatives | Alternatives to [rival] for [constraint] | "Cheaper alternatives to Descript for social video" |
| Head-to-head | [Your product] vs [rival] for [use case] | "Your product vs CapCut for product demo videos" |
| Job to be done | How do I [task] with AI | "How do I turn a webinar recording into ten short clips automatically" |
| Follow-up | A second message narrowing the first answer | "Which of those works best if we already use Canva" |
| Brand check | What is [your product] and who is it for | "What is your product and how much does it cost" |
The follow-up row is specific to chat. A buyer can add a constraint in a second message, and a product missing from the opening list can appear once they do, or drop out of it. Track the first answer and one scripted follow-up for your most important shortlist prompts.
Find out which AI engines recommend you
A free consultation on where you appear today across AI engines and search.
Schedule a Free ConsultationWhat is the difference between ChatGPT search answers and model answers?
A search answer is built from web pages ChatGPT retrieved for that prompt and shows citations; a model answer comes from what the model learned in training and shows none. They respond to different work. Search answers follow which pages can be found and quoted, while model answers follow how the wider web described your category before training.
OpenAI's help page says ChatGPT may search automatically when a question would benefit from current information, and that you can force a search from the tools menu by selecting Search, by typing / and choosing Search, or by regenerating a response with Search the web. When it searches, it rewrites your question into one or more targeted queries for its search providers, and the page lists Microsoft among them. That is why Bing rankings matter for ChatGPT search answers, and why OpenAI's crawler documentation warns that sites blocking OAI-SearchBot will not be shown in ChatGPT search answers.
- Search used: record it as yes when the answer shows citations or a Sources panel, no when it does not.
- Default run: send the prompt as is and let ChatGPT decide, because that is what your buyer sees.
- Forced search run: rerun with Search selected, which shows which pages ChatGPT can find for the prompt.
- Report the two apart: a mention rate that blends both modes hides which kind of work would move it.
How to track brand mentions in ChatGPT, step by step
A monthly ChatGPT tracking run
- Open a clean sessionStart a Temporary Chat and choose Unpersonalized before the first message, which OpenAI says does not use memory, custom instructions or plugins, or use a logged-out browser window.
- Run each prompt in default modePaste the prompt unchanged, then record whether search was used, who was named, in what order and which URLs were cited.
- Repeat in fresh chatsOpen a new Temporary Chat for every run so earlier answers in the same conversation do not steer later ones.
- Add a forced search runRerun with Search selected and log it as its own row, marked as forced.
- Open the cited URLsNote whether each source is your page, a rival's page or a third-party review, roundup, thread or video.
- Check the descriptionScore the answer against a one-page fact sheet of your category, audience, pricing model and core features.
- Match against GA4Compare the month's mention rate with sessions tagged utm_source=chatgpt.com and with signups from those sessions.
OpenAI's Temporary Chat FAQ confirms that unpersonalized temporary chats do not use memory, custom instructions or plugins, and that personalization cannot be changed once the chat starts. Choose unpersonalized every time.
What should a ChatGPT tracking sheet look like?
One row per answer, with the mode recorded on every row and absences logged as carefully as mentions. Here is what three rows look like for the hypothetical video editor above.
| Prompt and run | Mode | Named | Recommended | Competitors named | Cited URLs | Description issue |
|---|---|---|---|---|---|---|
| P01 shortlist, run 1 | Temporary Chat, search used | Yes, third of five | No | Rival A, Rival B, Rival D, Rival E | A roundup on a review site, Rival A's pricing page | None |
| P01 shortlist, run 2 | Temporary Chat, no search | No | No | Rival A, Rival B, Rival C | None | Not applicable |
| P05 brand check, run 1 | Temporary Chat, forced search | Yes | Not applicable | None | Your pricing page, an old launch post | Says there is no free plan |
A summary tab turns the rows into mention rate and recommendation rate per prompt group and per mode, plus share of voice against your competitor list; the calculation is in what share of voice in AI search is and how to calculate it. The description issue column is the one founders skip and regret. In the sample, the brand-check answer cites an old launch post, which tells you exactly which page to update.
How do you connect ChatGPT mentions to traffic and signups?
Match your monthly mention rate against ChatGPT referral sessions and the signups they produce. OpenAI's publishers FAQ says ChatGPT adds utm_source=chatgpt.com to referral links, so those sessions can be isolated in GA4 without guesswork. The GA4 setup guide has the channel group.
Clicks undercount influence. A buyer who reads a ChatGPT answer and searches your name on Google or Bing later shows up as branded search, not as a ChatGPT referral, so add "ChatGPT or another AI assistant" as an option in a required "How did you hear about us" field on signup and demo forms. If mentions are high and referral sessions near zero, check that the cited URLs are yours; if mentions are zero, start with why your brand does not show up in ChatGPT.
Which tools automate ChatGPT mention tracking?
Several SEO platforms now run custom prompts in ChatGPT on a schedule and report mentions, citations and share of voice. Ahrefs Brand Radar supports custom prompts with ChatGPT among its tracked platforms, and Semrush's AI Visibility Toolkit describes day-by-day prompt tracking on ChatGPT. They save hours, but they sample answers just as you do.
- How many times each prompt is run per check, and whether the tool reports a rate or a single position.
- Whether runs are logged out or personalized, and from which country.
- Whether it records if ChatGPT searched, and captures every cited URL rather than only your own.
- Whether you can export the raw answers, so a person can read them each month.
Ask a vendor those four questions before you sign. A tool that reports ChatGPT positions to one decimal place is reporting noise, given how rarely lists repeat. We compare the options in AI visibility tracking tools compared for startups; for a first look with no spend, use the one-hour check without paid tools.
How often should you check ChatGPT for brand mentions?
Monthly for the full prompt set, with weekly spot checks on two or three shortlist prompts around launches. Sources churn fast: SISTRIX's May 2026 study of 82,619 prompts over 17 weeks found ChatGPT replaced 74% of its cited sources each week, the highest of the platforms it tracked, and the median prompt did not have a single domain cited in all 17 weeks. Read three-month trends, then act. When the trend says you are losing shortlist prompts, that is the problem a ChatGPT SEO agency is hired to fix, and how ChatGPT decides which software to recommend explains the mechanism.
Frequently asked questions
Can you see how often ChatGPT mentions your brand?
Not directly. OpenAI does not offer site owners a report of brand mentions in ChatGPT answers. You can see clicks, because ChatGPT adds utm_source=chatgpt.com to referral links, but mentions have to be sampled by running your buyers' prompts yourself or through a tracking tool. Run each prompt several times in clean sessions and report the share of runs that name you, since a single answer is not representative.
Why does ChatGPT give different answers to the same prompt?
Answers are generated fresh each time, and ChatGPT may or may not search the web, may rewrite the search differently and may use memory and approximate location when personalization is on. SparkToro and Gumshoe found less than a 1% chance of ChatGPT returning the same brand list twice across repeated runs. That is why tracking uses repeated runs in unpersonalized sessions and reports rates instead of positions.
Should I track ChatGPT with search on or off?
Track both, and keep them apart. Run each prompt as a buyer would, letting ChatGPT decide whether to search, and record whether citations appeared. Then add a run with Search selected from the tools menu. Search answers show which of your pages ChatGPT can retrieve and quote; answers without search reflect what the model already believes about your category. They respond to different work, so a blended number hides the cause.
Do ads in ChatGPT affect brand mention tracking?
They should not affect the answers, but they can confuse a tracker. OpenAI began testing ads on the Free and Go tiers in February 2026 and says ads do not influence answers and are labeled as sponsored and visually separate. When you log runs, record sponsored units in their own column if you want competitive intelligence, and never count them as organic mentions or recommendations.
