How to track brand mentions in ChatGPT, Claude, Gemini and Perplexity: build a prompt panel, run it, and measure mention rate and share of voice.

On this page
In short
to track your brand mentions in AI, write a fixed list of questions your buyers ask (a prompt panel), run them regularly in ChatGPT, Claude, Gemini and Perplexity, and record whether your brand is named, in what position, next to which competitors and from which sources. Because answers change from one run to the next, measure rates over many questions and weeks, not single answers. You can do it by hand in a spreadsheet for a dozen prompts, or with a tracking tool for larger panels and daily runs.
Want to go further? Ask
Quick quiz
What should you work on first for your brand?
Sorank matches your answers
Test it for 3 days on your own site and see where you stand on Google, ChatGPT and the other AIs.
Start 3-day free trial3 days free, then stop whenever you want. 90-day results-or-refund guarantee.
Knowing how to track brand mentions in ChatGPT and other assistants has become a marketing basic: buyers ask them for recommendations, and the brands named get the shortlist. Assistants are used at massive scale; OpenAI's CEO announced more than 800 million weekly ChatGPT users in October 2025, as reported by TechCrunch. This guide gives a method, a template and the metrics that matter.
List 25 to 50 questions per market, phrased the way buyers talk, across the buying journey:
| Type | Example |
|---|---|
| Category discovery | What are the best invoicing tools for freelancers? |
| Use case | Which CRM works for a five-person team using Gmail? |
| Comparison | Brand A or Brand B for a small agency? |
| Alternatives | What are the alternatives to Brand A? |
| Brand facts | How much does [your brand] cost? Is [your brand] reliable? |
| Local | Best accountant for startups in Lyon? |
Sorank's AI prompt research suggests questions from your market, and our prompt library gives examples.
Run each prompt in each assistant you care about, in a clean session (logged out or in a temporary chat, without memory), in the language and country of your market. Repeat on a fixed schedule: weekly for a manual panel, daily with a tool. Assistants with web search (ChatGPT search, Perplexity, Gemini) reflect current sources; answers without search reflect training data.
Note
The same prompt can give different answers minutes apart. That is why one test proves nothing: a mention rate over 25 prompts, repeated each week, gives a reliable trend.
| Field | What to note |
|---|---|
| Date, assistant, prompt | Context of the run |
| Mentioned | Yes or no |
| Position | First, in the list, or in passing |
| Competitors named | Who appears with or instead of you |
| Sources cited | Pages and sites linked or quoted |
| Accuracy | Correct or wrong facts about your brand |

| Approach | Good for | Limits |
|---|---|---|
| Spreadsheet, by hand | 10 to 20 prompts, a first diagnosis | Time-consuming, irregular, small sample |
| Tracking tool | 25 to 150+ prompts, daily runs, several assistants | Subscription cost |
Sorank's AI mention tracking runs 25 to 150 prompts depending on the plan in ChatGPT, Claude and Gemini, plus Perplexity and Grok from the Pro plan, with a GEO leaderboard against competitors. Dedicated pages cover the ChatGPT rank tracker and the Claude rank tracker. Our ranking of AI visibility tools compares the options.
Tip
Add referral tracking in analytics too: visits from chatgpt.com, perplexity.ai and gemini.google.com show the clicks that mentions bring. Our guide to measuring AI citations in Google Analytics explains the setup.
Watch out
Do not judge your AI visibility from your own logged-in account. Personalization and chat memory can make an assistant mention your brand because you asked about it before. Use clean sessions or a tool.
Tracking AI mentions means a fixed panel of buyer questions, run regularly, measured as rates. Start by hand to learn, then automate. See the Sorank plans, with a 3-day free trial; for an overview of where assistants find their sources, our reference paper on GEO and the Pew study on AI summaries are good starting points.
Yes, by hand: write 10 to 20 buyer questions, run them each week in a clean ChatGPT session and note in a spreadsheet whether your brand appears and who is named instead. It takes time and gives a small sample; tools automate larger panels and daily runs.
Language models generate each answer anew, and assistants with web search also depend on which sources they retrieve. Answers therefore vary between runs. Measure mention rates over many prompts and weeks instead of reacting to a single answer.
Start with 25 questions that cover your main offers and buying stages. Brands with several products, markets or languages often track 50 to 150. Keep the list stable so trends remain comparable.
THEY TRIED SORANK BEFORE YOU
