# How to track whether ChatGPT mentions your brand: a step-by-step guide

> Track whether ChatGPT mentions your brand: the manual method with a tracking sheet, why it breaks down, what a tool should do, and how to measure traffic.

**Published:** 2026-11-04 · **Category:** GEO

[Version HTML](https://www.citeme.io/ressources/how-to-track-chatgpt-brand-mentions)

---

To track whether ChatGPT mentions your brand, write down 20 to 50 questions your buyers really ask, run each one several times in a neutral ChatGPT session, and record five things per answer: whether your brand is mentioned, whether it is recommended, its position, the sources ChatGPT cites, and whether what it says about you is accurate. A spreadsheet is enough for a first snapshot. Once you need weekly numbers, competitor comparisons or more than one AI engine, the manual method stops scaling and a tracking tool takes over.

**In short**: ChatGPT brand tracking is a sample, not a lookup. One answer proves nothing because the same question returns different answers. You measure a mention rate across many runs and many prompts, then you follow it over time.

## What does tracking ChatGPT brand mentions actually measure?

It measures how your brand shows up inside the answers ChatGPT generates for questions about your market. There is no ranking page to check. Either the answer names you or it does not, and when it does, the way it names you matters as much as the mention itself.

Five signals cover most of what a marketing team needs:

-   Mention: does your brand appear anywhere in the answer?
-   Recommendation: is it presented as an option the user should consider, or only cited in passing (or as a cautionary example)?
-   Position: is it the first brand named, the third, or the last in a list of eight?
-   Sources: which pages and domains does ChatGPT link to when it answers? Are any of them yours?
-   Accuracy: are your prices, features, locations and positioning correct, or is ChatGPT repeating something outdated?

Add one more column for competitors named in the same answer. That is where the useful insight usually sits: a brand that is never mentioned on "best X for Y" questions, while two rivals appear every time, knows exactly which battle it is losing.

> **In short**: track mention, recommendation, position, sources and accuracy, and always note which competitors appear next to you.

## How do you track ChatGPT mentions manually?

The manual method costs nothing but time, and it is the right way to start. It teaches you how ChatGPT talks about your category before you pay for anything.

### Step 1: write 20 to 50 real buyer questions

Use the questions prospects actually ask, not the keywords you want to rank for. Good places to find them: sales call notes, support tickets, the question-shaped queries in Google Search Console, and threads on Reddit or industry forums. Mix four types:

-   Category questions: "What is the best payroll software for a 20-person company?"
-   Comparison questions: "Brand A vs Brand B for small agencies?"
-   Problem questions: "How do I stop invoices from being paid late?"
-   Brand questions: "Is YourBrand reliable?" or "What does YourBrand cost?"

Category and problem questions tell you whether ChatGPT thinks of you at all. Brand questions tell you whether what it says is correct.

### Step 2: use a neutral session

ChatGPT adapts to the person asking. If you test from your own account, with months of conversations about your company and memory switched on, the model may mention you because it knows you care. Test while logged out, or use a Temporary Chat, which does not draw on memory. Turn off custom instructions. If your buyers are in another country, keep in mind that location can change which local brands come up.

### Step 3: run each question several times

Run every prompt at least three times, each in a new conversation. Five is better on the prompts that matter most. You are not looking for "the" answer. You are looking for how often your brand comes back.

### Step 4: note whether ChatGPT searched the web

ChatGPT answers some questions from what the model learned during training and others by running a web search first. When it searches, the answer shows linked sources. The distinction matters for what you do next. An absence in answers without search points to your brand's footprint across the web over time. An absence in answers with search points to the pages ChatGPT retrieved that day, which you can study and sometimes join.

### Step 5: log everything in one sheet

Use one row per run. Here is a template you can copy. The two rows are fictional examples to show the format, with "YourBrand" standing in for your company.

| Date | Prompt | Run | Web search | Mentioned | Recommended | Position | Competitors named | Sources cited | Accuracy issue |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 2026-11-04 | Best payroll software for a 20-person company | 1 | Yes | Yes | Yes | 3 of 6 | Brand A, Brand B | 2 review sites, 1 vendor blog | Old price quoted |
| 2026-11-04 | Best payroll software for a 20-person company | 2 | No | No | No | None | Brand A, Brand C | None shown | None |

### Step 6: turn the rows into two numbers

Mention rate is the number of runs that mention you divided by the total number of runs. Share of voice is your mentions divided by all brand mentions across those runs, yours and your competitors'. Both numbers mean something only when you compare them over time, or against a named rival on the same prompt set.

> **In short**: real buyer questions, a neutral session, several runs per prompt, one row per run, then mention rate and share of voice.

## Why does the manual method break down so quickly?

It works for a first look, and a weekly check by hand is manageable on a short list of 10 prompts or so. Once the list reaches the 20 to 50 prompts a real measurement needs, it does not hold up week after week, for four reasons.

### The answers are not deterministic

Ask ChatGPT the same question twice and you can get two different lists of brands. That is how language models generate text, not a bug. A single screenshot of ChatGPT recommending you, or ignoring you, is an anecdote. You need enough runs for the rate to settle, and that number grows with the number of prompts you care about.

### Personalization leaks into every test

Memory, chat history, custom instructions and location all shape answers. Even with a clean session, two people on the same team rarely run tests under identical conditions, so their results drift apart for reasons that have nothing to do with your visibility.

### ChatGPT is not one engine, and it is not the only one

Answers change with the model behind ChatGPT, with whether a web search ran, and with each model update. Outside ChatGPT, your buyers also ask Perplexity, Gemini, Claude and Google AI Overviews, and those engines lean on different sources. In our own tracking panel at Citeme (our project for citeme.io, last 30 days to 3 November 2026), 15% of the sources cited by Google AI Overviews were social platforms such as YouTube, Reddit and LinkedIn, against 2% or less for ChatGPT, Gemini, Claude and Perplexity. That is one panel, not the whole web, but it shows why a result on ChatGPT says little about the others.

### The volume gets out of hand

Do the arithmetic. 30 prompts with 5 runs each is 150 answers to read and log for one engine. Track four engines and it is 600 answers. Repeat it every week and someone on your team now has a part-time job copying text into a spreadsheet, with plenty of room for coding errors along the way.

> **In short**: random answers, personalization, several models and several engines, multiplied by volume. Manual tracking gives you a snapshot, never a trend you can trust.

## What should an automated ChatGPT tracking tool do?

The tool should do what you did by hand, at a volume and regularity you cannot match. Before you pick one, check that it covers these points:

-   Repeated runs on a schedule, daily or weekly, on the same prompt set, so the mention rate is measured and not guessed.
-   Several engines from one dashboard, at least the ones your buyers use.
-   Share of voice against competitors you name, prompt by prompt.
-   The sources cited in each answer, down to the URL, so you know which pages to win a place on.
-   The text of the answer, so you can check accuracy and tone yourself.
-   Alerts when you gain or lose a citation, and a history that goes back far enough to see a trend.

### Citeme (our product)

Citeme is our product, so read this part with that in mind. It tracks your prompts every day across up to 10 engines: ChatGPT, Claude, Gemini, Perplexity, Grok, Google AI Overviews, Siri AI, DeepSeek, Meta AI and Mistral. What we built it around is prompt-level resolution: for every citation you win or lose, it shows the exact prompt where it happened, the answer and the sources the engine used.

![Citeme prompt tracking table with visibility, average position, citing AI engines, markets, trend and last run for each prompt](https://cdn.sanity.io/images/xc5dg2px/production/c6bafa784558c75f5ec6dafd88efd06a8540025c-2880x2048.webp)

_Each tracked prompt gets its own row with its visibility, its average position, the engines that cite the brand and the date of the last run._

| Plan | Price per month (excl. tax) | Prompts tracked daily | Engines | Brands |
| --- | --- | --- | --- | --- |
| Starter | $69, or $59 billed annually | 50 | 3 of your choice among 10 | 1 |
| Growth | $149 | 125 | 4 (ChatGPT, Perplexity, Claude, Gemini) | 3 |
| Pro | $349 | 300 | 4 | 10 |

Each extra engine costs $25 per month, users are unlimited on every plan, and you can cancel at any time. Slack and email alerts start on Growth. You can try it for 3 days for $1, with a payment method required. Full details are on the [pricing page](https://www.citeme.io/pricing).

### Other tools

Citeme is not the only option. Profound, Otterly and Peec AI are among the tools built for the same job, and broader SEO suites such as Semrush have added AI visibility features. We compare ten of them in our guide to [the best AI visibility tools in 2026](https://www.citeme.io/ressources/the-10-best-ai-visibility-tools-in-2026-track-your-brand-across-every-ai-search-engine).

![Bar chart of the price per tracked prompt across Citeme, Peec, Meteoria, Otterly and Profound plans, from $1.16 to $3.99](https://cdn.sanity.io/images/xc5dg2px/production/5b5d87dedd465bc5d1fe6fe6040f585b20e53fff-2448x1786.png)

_Price per tracked prompt for Citeme, Peec, Meteoria, Otterly and Profound plans, computed from public prices checked on 27 October 2026 (our chart)._

> **In short**: a tracking tool should rerun your prompts on a schedule, across engines, with competitors, sources and alerts. Citeme does this from $69 a month and shows the prompt behind each citation won or lost.

## How do you measure the traffic ChatGPT sends you?

Mentions are half the picture. The other half is what happens when someone clicks a link in ChatGPT and lands on your site.

Links from ChatGPT usually carry a `utm_source=chatgpt.com` parameter, which shows up in Google Analytics as a referral. Do not treat that label as complete. Our own research report ([Who read this page?](https://www.citeme.io/research/who-read-this-page)) found that ChatGPT's links went out with no label for five days in October 2026, and on ten client sites the number of tagged ChatGPT clicks per day fell by 60% during that window. Any report built only on the tag undercounts.

Two more sources fill the gap. Your server logs record visits from ChatGPT-User, the agent OpenAI uses when a user's request makes ChatGPT open a page. And a dedicated layer can join the two: Citeme's [AI analytics](https://www.citeme.io/features/analytics) separates bots from humans who arrived from an AI answer, and [AI Impact](https://www.citeme.io/features/ai-impact) ties those visits to the most likely prompt and, with a connected CRM, to the deals they produced.

> **In short**: use the chatgpt.com referral as a starting point, check it against server logs, and expect the tag to miss some clicks.

## Can ChatGPT read your site in the first place?

Before you worry about mentions, make sure you are not blocking the crawlers that feed ChatGPT. OpenAI documents its user agents on its [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) page, and it states that each setting is independent of the others.

| User agent | What OpenAI says it does | If you block it |
| --- | --- | --- |
| OAI-SearchBot | Surfaces websites in search results in ChatGPT's search features | Your site is not shown in ChatGPT search answers. OpenAI says a robots.txt change can take about 24 hours to apply |
| GPTBot | Collects content used to train OpenAI's foundation models | Your content is not used for training. ChatGPT search is not affected |
| ChatGPT-User | Visits pages for certain user actions in ChatGPT and Custom GPTs | OpenAI says robots.txt rules may not apply, since a user starts the action. A firewall rule can still stop these visits |

A firewall or CDN rule can block these agents even when robots.txt allows them, so check both. Test your robots.txt against each of these user agents, not only GPTBot.

![Citeme AI bots dashboard showing 15,809 bot visits from 12 AI bots over 30 days, crawl activity per bot and hourly peaks](https://cdn.sanity.io/images/xc5dg2px/production/1a6196e675d0de9acbc6723a7a06d42570ca07b1-4446x974.png)

_A real screen of the Citeme platform: the AI bots that crawled the site over the last 30 days, ChatGPT among them, with none blocked by robots.txt._

> **In short**: allow OAI-SearchBot if you want to appear in ChatGPT search. GPTBot is a separate training decision.

## What do you do with the results?

Tracking only pays off if it changes what you work on. Read your sheet, or your dashboard, by pattern rather than by prompt.

If you are missing from answers that use web search, open the sources ChatGPT cited. They are often comparison articles, review sites and forum threads. Those pages are your shortlist: get listed, get reviewed, or publish a page that answers the question more directly than they do.

If you are missing from answers without search, the gap is your footprint across the web. That takes longer and depends on independent mentions accumulating over months.

If ChatGPT gets your facts wrong, fix the source of the error first. Update your own pages, then the third-party profiles that still show the old information.

Our guide on [how to rank in ChatGPT](https://www.citeme.io/ressources/how-to-rank-in-chatgpt-the-complete-guide-to-getting-recommended) covers those fixes step by step. If you want a starting measurement across all ten engines, the [AI visibility checker](https://www.citeme.io/tools/ai-visibility-checker) runs inside the Citeme app during the 3-day trial for $1.

> **In short**: absent with search means work on the cited pages, absent without search means build mentions over time, and wrong facts mean fixing the source.

## FAQ

### Does ChatGPT give everyone the same answer about my brand?

No. The same question can return different brands from one conversation to the next, because the model generates each answer fresh. Memory, chat history, custom instructions, location, the model in use and whether a web search ran all add variation. That is why a single test means little. Measure how often your brand appears across several runs of each prompt, in a neutral session, and follow that rate over time.

### How often should I check whether ChatGPT mentions my brand?

Weekly is a reasonable rhythm for a manual check, as long as the list stays short (about 10 prompts). More often than that, or on a longer list, the work becomes hard to sustain by hand. Daily tracking makes sense when you are actively publishing changes and want to see their effect, or when competitors move fast in your category. At that frequency, a tool is the only realistic option.

### Can I see ChatGPT mentions in Google Analytics?

Not the mentions. Google Analytics only sees visits, so it shows the people who clicked a link in ChatGPT, usually as a chatgpt.com referral. Someone who read a recommendation and later typed your name into Google shows up as branded search or direct traffic. And the referral label itself is not reliable: in our own research, ChatGPT links went out untagged for five days in October 2026. Mentions have to be measured inside the answers.

### Is there a free way to track ChatGPT mentions?

Yes, the manual method in this guide: a list of buyer questions, a neutral session, several runs per prompt and a spreadsheet. It costs time instead of money. Most tracking tools are paid. Citeme has no free plan: its checker runs during a 3-day trial for $1, with a payment method required.

### Does blocking GPTBot stop ChatGPT from mentioning my brand?

Not directly. According to OpenAI, GPTBot is about training future models, while OAI-SearchBot controls whether your site can appear in ChatGPT search answers. Blocking GPTBot does not remove you from search answers, and it does not erase what earlier models already learned. ChatGPT can also mention your brand based on what other sites say about you, whatever your own robots.txt allows.

## Sources

- [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) · OpenAI
- [Who read this page? (research, October 2026)](https://citeme.io/research/who-read-this-page) · Citeme
- [Pricing](https://citeme.io/pricing) · Citeme