Ask ChatGPT about your industry and you'll get an answer with no dashboard, no position number, and no obvious way to tell whether your brand showed up at all. Tracking brand mentions in AI search means measuring how often you appear across repeated runs, then watching that rate instead of chasing a rank that doesn't exist.
The metric is mention rate, not rank
A keyword has one Google ranking. An AI answer doesn't. Ask the same question twice and the wording changes, and whether your brand gets named can flip between runs.
A study from SparkToro and Gumshoe in early 2026 had 600 people run 2,961 prompts through ChatGPT, Claude, and Google's AI. The chance that the same brand list appeared twice for an identical prompt was under one percent. Matching the order was rarer still.
That finding means any tool showing you a position is measuring noise. The stable number is frequency: across many runs, how often does your brand appear? In narrow categories the top brands showed up in 55 to 77 percent of responses regardless of how the prompt was phrased. Name your metric mention rate and the rest of the work falls into place.
Build a baseline by hand this week
You don't need a subscription to start. You need an hour and a spreadsheet.
- Write twenty questions your customers would actually ask. Comparison questions work well, like which of two products is better for a given job.
- Run each one through ChatGPT, Perplexity, and Google's AI summaries.
- For each answer, note whether your brand appeared, roughly where in the response, and which competitors sat beside it.
- Repeat the same set on three different days. One result can be a fluke.
- Build a table of mention rate per question.
That table is your baseline. Rebuild it a month later and the gaps tell you what changed. Watch the ratio across twenty questions, not one prompt in isolation. Why not just check once and move on? Because a single answer is a coin flip, not a measurement.
The three engines don't behave alike
Each engine surfaces mentions differently, so the way you track them shifts.
ChatGPT pulls from training data and from live pages when it browses, but it attaches source links in only a minority of answers. So with ChatGPT you track the mention itself: is your name there, in what context, with what tone.
Perplexity runs a live crawl and lists sources on nearly every answer, around twenty-odd citations per response. That makes it the easiest engine to check and the one where an AI mention can actually turn into a click. If you want AI traffic in your analytics, start here.
Google's AI summaries draw on the regular Google index and link their sources inline. Mentions there are tangled into the traffic you already see in Search Console, which makes them harder to isolate but easier to tie to demand you understand.
Reading a number without knowing these differences is how a team concludes that ChatGPT mentions rose while traffic stayed flat. That isn't a mystery. ChatGPT simply didn't attach links.
Turning a low mention rate around
If your baseline shows you're missing, five levers usually explain it.
First, write in clusters. Covering several related questions around one topic means more chances to be picked up when the model breaks a broad question into sub-questions.
Second, earn mentions on sites with domain authority. A large share of ChatGPT's citations trace back to high-authority pages, so a mention on a respected site moves the needle more than ten posts on your own.
Third, refresh key pages on a schedule. Recently updated content gets cited more than something untouched for a year.
Fourth, answer up front. Put a direct response near the top of the page and back it with structured data like FAQ markup so the model can lift it cleanly.
Fifth, target real queries. Take the exact phrasing a customer would type into an AI assistant and put it into your page titles and subheadings.
When to pay for a tool
Manual tracking holds up to about twenty questions. Past that, the repetition eats your week and you want automation. Paid tools generally start around twenty to thirty dollars a month and run your prompt set across several platforms, logging mention rate and the cited sources.
When you compare them, look at how each one calculates mention rate rather than at the size of its rank display. Tools that lead with ranking carry the noise problem described above. The ones that sample repeatedly by default and show both mention rate and citation sources are the useful ones, because a single reading of a number that changes every run tells you nothing.
Your first move
This week, settle on twenty questions and run them across the three engines to build the baseline table. Then pick the five questions where you don't appear and check whether your site has any page that answers them. If it doesn't, that gap is the first thing to fill. Repeat the measurement monthly, and you'll be able to see which content changes actually moved your mention rate.






