How to track brand mentions in ChatGPT, step by step

To track brand mentions in ChatGPT, you ask it the same questions your buyers ask, on a schedule, and count how often the answers name you, in what position, in what tone and next to which competitors. ChatGPT has no report for brands and no fixed ranking, so the measurement is yours to build. This guide covers the prompts to use, what to count, how often to check and the free and paid tools that do the work.

We build one of those tools, Mencoro, and track our own brand this way. Where it helps, we show our own numbers.

Why ChatGPT mentions are hard to track

Google gives you Search Console. ChatGPT gives you nothing comparable: OpenAI does not tell brands when or how its answers name them. Three things make the job harder than classic rank tracking:

• There is no position one. ChatGPT writes an answer and names a few brands. You measure whether you are named and how, not a slot in a list of links.
• Answers change between runs. The same question can name you today and leave you out tomorrow, and answers shift again when OpenAI updates the model.
• Answers are personal. Memory, custom instructions, language and country change what each user sees, so your own logged-in account is a poor sample.

Your analytics only catch the end of the story. Visits from ChatGPT show up as AI referral traffic, tagged utm_source=chatgpt.com, but most people read the answer and never click. A brand can be recommended a thousand times and see a handful of sessions.

Step 1: write the prompts your buyers ask

The prompt set is to ChatGPT tracking what a keyword list is to SEO. Write 20 to 50 questions the way a buyer would type them, and group them by intent:

• Category: “best CRM for a small agency”, “top project management tools”.
• Comparison: “HubSpot vs Pipedrive”, “alternatives to Asana for a remote team”.
• Problem: “how do I stop losing leads from my website form”.
• Brand: “is [your brand] good for agencies”, “[your brand] pricing”.

Keep branded and generic prompts in separate groups and report them apart. A question that already contains your name will mention you almost every time, and mixing it in inflates the number that matters: how often ChatGPT brings you up when nobody asked about you. Once the set is fixed, keep it fixed. If you add prompts, add them as a new group so the trend stays comparable.

Step 2: decide what counts as a mention

Write down every name your brand goes by, your domain and the misspellings people use, and do the same for the three to five competitors you compare against. Then decide what you count. Not every AI brand mention is worth the same:

• Recommendation: ChatGPT tells the user to pick you.
• Comparison: you are weighed against another brand.
• List entry: you are one of several names in a list.
• Example or reference: you are named in passing.
• Citation: a link to your site among the sources, with or without your name in the text. See AI citations.

Log the type next to every mention. A month of passing references and a month of recommendations give the same count and mean very different things.

Step 3: run a manual baseline

Before you buy anything, run the set once by hand. Open a temporary chat, which skips your memory and custom instructions, make sure search is on, and ask each prompt as written. For each answer, log in a spreadsheet:

Twenty prompts take about an hour. If you sell in several countries, ChatGPT answers from where you are, so a manual check only shows your own market. Our free ChatGPT rank checker runs up to three prompts and shows your position in each answer, and the AI visibility checker compares you with your competitors in ChatGPT, Gemini and Perplexity, both with no sign-up.

Step 4: repeat the prompts on a schedule

One run is a snapshot. Because ChatGPT can answer the same question differently each time, the useful figure is how often you are named over many runs. Check your most valuable prompts daily and the rest weekly, and run each prompt two or three times per check: named in two of three passes counts as two thirds, which smooths out answers that vary.

This is where manual tracking breaks. Fifty prompts, three passes, once a week is 600 answers a month to read and log. A ChatGPT rank tracker runs them for you, records the full answer and its sources, and keeps the history.

Step 5: measure the numbers that matter

Four numbers cover most of what you need:

• Mention rate. Answers that name you divided by answers checked, times 100. See mention rate.
• Position. Where you are first named among the brands in each answer, averaged. Lower is better.
• Share of voice. Your mentions against your competitors’ for the same prompts. See share of voice.
• Sentiment. The tone of each mention, scored. See brand sentiment in AI.

Worked example: 30 prompts checked weekly for four weeks are 120 answers. You are named in 36, a mention rate of 30%. Your main competitor is named in 60, or 50%. The prompts behind the answers where it is named and you are not are your to-do list.

Step 6: compare engines and countries

ChatGPT is one engine of several, and the same brand can look very different on each. In our own project for mencoro.com, between 28 October and 10 November 2026, Mencoro was named in 13 of 174 ChatGPT answers (7%), against 21 of 166 in Perplexity (13%) and 18 of 129 Google AI Overviews (14%). In the United States it was 4%, well below Spain.

The same project shows why step 1 matters. Our overall rate rose from about 1% to 11% in two weeks, but almost all of the rise came from new long-tail prompts that describe our own features. On the generic category questions we started with, Mencoro was named in 1 answer in 60. Had we mixed the two groups, we would be celebrating a change in the prompt set, not in ChatGPT.

So track ChatGPT next to the engines your buyers use, country by country, and read each one on its own. If you also want to see how AI describes you across engines, that is the job of AI brand monitoring.

Step 7: act on the gaps

Tracking is only useful if it tells you what to do. Start with the prompts where a competitor is named and you are not, and read the sources ChatGPT cites for them. Those pages are the shortlist: comparison articles, reviews, directories and forums that ChatGPT already trusts. Earning a place on them usually moves the answer more than another page on your own site.

Then improve your own pages for the questions you lose: a direct answer up top, clear facts, and the same description of your brand everywhere. Our guide on how to rank on ChatGPT covers those levers step by step. Re-run the set and watch whether the gap closes.

Manual tracking, a free checker or a tracker

Mencoro tracks ChatGPT next to Perplexity, Claude, Google AI Overview and AI Mode in one project, classifies every mention into five types, scores its tone and weighs your share of voice against competitors, from €15 a month. For other options, see our comparison of the best ChatGPT rank trackers, and the methodology for how each figure is calculated.

Common mistakes

• Checking from your own account. Memory and past chats bias the answer towards you. Use a temporary chat or a tracker.
• Changing the prompts every month. The trend only means something with a fixed set.
• Counting every mention the same. A recommendation and a passing reference are not equal; record the type.
• Reading one run. Answers vary. Look at the rate over weeks, not at a single screenshot.
• Judging by traffic. Most mentions never send a visit, so referral traffic underestimates your visibility.