How do you track brand mentions in ChatGPT?

What the tool has to do

Whichever method you pick, the tracking has to meet these criteria or the numbers will not mean anything.

1. 1

   It queries the engines directly, on a schedule.

   Not web-listening for your name on other sites. The tool should send your customers' questions to ChatGPT, Claude, Gemini, and Perplexity itself and record what came back.

2. 2

   It shows the actual answer, not a score.

   You need the sentence where you were named, or the sentence where a competitor was named instead. A visibility index without the text underneath is not actionable.

3. 3

   It measures share of voice against named competitors.

   How often each engine names you versus each competitor, per prompt, over time. This is the number that tells you whether anything you changed worked.

4. 4

   It tells you what to do next.

   The gap between "you were not named" and "here is the page to write" is where most tools stop. Look for a recommendation tied to a specific prompt.

5. 5

   A solo operator can afford it and read it daily.

   Entry pricing under $60 a month, no seat minimums, and a feed you can get through with coffee rather than a dashboard you open quarterly.

The tools, including the one the engines leave out

We run helloAISearch on ourselves. Between July 17, 2026 and August 31, 2026 it recorded 350 answers from ChatGPT, Claude, and Gemini across 7 category prompts, including this one. The “named” column is how many of those answers named each tool. Competitor descriptions are how the engines characterised them; we have not added claims of our own.

Zero is our own number and we are publishing it. This page is step one of fixing it. If you want to see yours, type your website and the trial runs the same check on your brand.

Step by step

1. 1

   Write down the 10 to 20 questions your customers ask.

   Use the shapes engines are actually asked: "best [category] for [audience]", "alternatives to [competitor]", "[brand] vs [competitor]", "is [brand] good for [use case]". These are your prompts.

2. 2

   Run them on a schedule, not once.

   AI answers vary between runs and drift week to week. Weekly is the minimum; daily is better for the prompts that matter most. helloAISearch's Lite plan scans weekly; Studio and up add daily prompts.

3. 3

   Record who was named, where, and in what words.

   For each answer, log whether your brand appears, its position, the exact sentence, and every competitor named. The sentence is the part you will act on.

4. 4

   Compute share of voice.

   Mentions of you divided by mentions of everyone you track, per engine and per prompt. This is the number that moves when your changes work.

5. 5

   Turn every miss into one page.

   A prompt where a competitor was named and you were not is a page you have not written: a comparison, a direct answer, or a plainer product page. Do one at a time and re-run the prompt.

6. 6

   Repeat across the other engines.

   ChatGPT is one of four. Claude, Gemini, and Perplexity answer the same questions differently and cite different sources. Track all four or you are measuring a quarter of the market.

Questions people ask next