How to Track Your Brand's Visibility in ChatGPT

ChatGPT is where most buyers' AI habit started, which makes it the engine marketing teams ask about first. The question is simple — "when people ask ChatGPT about our category, do we come up?" — and the honest answer requires more care than a screenshot of one good response.

Here's the method that produces data you can defend, whether you run it by hand or with software.

Step 1: build a prompt set from real buying questions

Don't test the questions you wish buyers asked; test the ones they do. The reliable categories: best-of ("best [category] software for [segment]"), comparisons ("[rival A] vs [rival B]"), alternatives ("alternatives to [market leader]"), pricing ("how much does [category] cost"), and problem-first questions ("how do I [job your product does]"). Fifteen to twenty-five prompts covers a category's core.

One design rule that matters: most of your prompts should not contain your brand name. The measurement is whether ChatGPT surfaces you unprompted — seeding your name into every question is how teams accidentally manufacture good news.

Step 2: ask more than once, in clean sessions

Two mechanics protect the data. First, use fresh chats — ChatGPT's memory of a conversation (and, if enabled, of you) contaminates later answers, and a logged-in session where you've discussed your own company is hopelessly biased. Test from a clean state.

Second, repeat each prompt. Identical questions produce different brand lists between askings often enough that single-pass checks are closer to coin flips than measurements. Three runs per prompt is the floor we use; "recommended in 2 of 3 runs" is a fact you can put in a deck.

Step 3: score three things, not one

For each answer record whether you were mentioned (named at all), recommended (actively suggested), and cited (your site referenced as a source). The distinction pays off immediately in diagnosis: lots of mentions but few recommendations means ChatGPT knows you but isn't sold — usually a positioning and review-signal problem. No mentions at all is a different disease with different medicine.

Log competitors in the same pass. In one category we baselined, the market leader was mentioned in roughly half of ChatGPT's answers — while a challenger brand assumed ChatGPT ignored their space entirely and was wrong. You don't know which story is yours until you look.

Don't stop at one engine

The catch with ChatGPT-only tracking: engines disagree, a lot. In our four-engine runs, the same brand set that ChatGPT mentioned in 49% of answers showed up in 81% of Grok's and 43% of Perplexity's. If your buyers research on Perplexity and you optimized for what ChatGPT says, you fixed the wrong thing. Track ChatGPT first if you must start somewhere — then widen. The Treyci baseline runs your prompt set across ChatGPT, Perplexity, Gemini, and Grok with three runs each, from $99; the methodology is public if you'd rather build it yourself.

Frequently asked questions

Keep reading

• What Does AI Say About Your Brand? How to Find Out — and Track It
• LLM Rank Tracking, Explained: Why It's Nothing Like Google Rank Tracking
• How to Monitor Brand Mentions in AI Search (Without Fooling Yourself)