How to track brand mentions in ChatGPT (the method that holds up)

To track whether ChatGPT mentions a brand, you need three things: a fixed set of buyer questions, asked repeatedly on a schedule, with every answer recorded — whether the brand was mentioned, which competitors were, and which websites were cited. Then you report the rate across all those answers, not any single one.

That's the whole method. The rest of this guide is how to do each step properly, and where doing it by hand stops being realistic for an agency.

Why one check tells you almost nothing

ChatGPT doesn't have a fixed ranking. Ask the same question twice and you'll often get a different list of brands, in a different order. In a study published by SparkToro and Gumshoe in January 2026, ChatGPT and Google's AI returned the same list of brands less than once in a hundred runs of the same prompt — and the same list in the same order roughly once in a thousand.

What was stable was how often a brand appeared. In that study, one hospital appeared in 69 of 71 ChatGPT answers to a cancer-care question but was the first name in only 25 of them. That's the number worth tracking: the share of answers that mention the brand. A screenshot of one answer shows one draw from that distribution. We explain the mechanics in why ChatGPT gives different answers every time.

Step 1: Write a fixed prompt set

Write the questions a real buyer would ask before they know the brand — not "what is Northwind Bikes?", but "best e-bike shop in Hamburg". About 20 prompts per client is a good default, spread across intents:

• Best options: "best [category] for [audience]"
• Alternatives: "alternatives to [market leader]"
• Comparisons: "[competitor] vs [competitor]"
• Problems and use cases: "how do I [job the product does]"
• Local, if relevant: "[service] near [city]"

Freeze the set. If you change prompts every month, you can't compare months. Our guide to choosing prompts for AI visibility tracking goes deeper.

Step 2: Decide what counts as a mention

Agree on this before you start, or your numbers won't be consistent:

• Brand names and aliases. Include the company name, product names, common abbreviations and older names after a rebrand.
• Mentioned vs cited. A mention is the brand named in the answer text. A citation is the brand's website among the sources the answer links to. Track them separately — a brand can be cited without being named, and named without being cited.
• Competitors. Pick 3–5 real competitors and track them the same way. Without them, a mention rate has no context.

Step 3: Ask each prompt repeatedly

One answer per prompt per month is not enough to see change. More answers per period means a more reliable rate. Two practical options:

• Weekly: ask every prompt once a week. With 20 prompts that's about 80 ChatGPT answers per month — enough to see a meaningful pattern, especially combined with other engines.
• Daily: better for detecting short-term movement on a few important prompts; expensive by hand.

How many answers you need to call a change real is covered in how often to track AI visibility.

Use a logged-out or clean session if you're checking manually, and keep the country and language consistent. Personal history and location change what ChatGPT says.

Step 4: Record every answer

A spreadsheet with one row per answer works:

Keep the full answer. The rate is what you report; the answer is what you show when someone asks why.

Step 5: Report rates, not results

At the end of the month, calculate:

• Mention rate: answers mentioning the brand ÷ all answers. "Mentioned in 31 of 100 answers."
• Competitor rates on the same answers, so the client sees who is mentioned instead.
• Citation rate: answers citing the brand's website ÷ all answers.
• Prompts where the brand never appears but competitors do. These are the most actionable lines in the report; why ChatGPT recommends a competitor helps you explain them.

Definitions and edge cases are in how to measure AI visibility.

The maths that ends manual tracking

For one client, the method above is a few hours a month. For an agency it isn't:

At even two minutes per answer, that's 120 hours a month — before anyone writes a report. This is the point where agencies either stop tracking, start sending screenshots (which prove nothing), or automate.

Automating it

Automated tracking does the same five steps on a schedule. Two methods exist: tools that call the AI providers' APIs with web search enabled, and tools that collect answers from the consumer apps. API answers can differ from what one person sees in the ChatGPT app — the app adds its own instructions and features — so whichever you use, report patterns over many answers rather than claiming to reproduce one person's screen.

Ranxii's ChatGPT visibility tracking uses the official APIs: it asks each client's prompts on ChatGPT, Gemini and Perplexity every week, records mentions, competitors and cited sources for every answer, and keeps the full text so each number opens the answers behind it.

Bonus: see the traffic, too

Mentions are half the picture. OpenAI says ChatGPT adds utm_source=chatgpt.com to links in its answers, so visits it sends show up as a source in Google Analytics and similar tools. Put that next to the mention rate in your report: one shows whether the brand is recommended, the other whether people click.

Sources

• SparkToro: AIs are highly inconsistent when recommending brands or products (January 2026)
• OpenAI Help Center: Publishers and developers FAQ (utm_source=chatgpt.com)