How to track whether ChatGPT mentions your brand
Start with a fixed list of questions and a spreadsheet. Automate only once you know what you are counting.
To track whether ChatGPT mentions your brand, ask it the same fixed set of buyer questions on a schedule, record for each answer whether it names your brand and whether it links to your site, and compare those records over time. One answer tells you almost nothing, because answers vary from run to run. A repeated, recorded set of answers tells you a lot.
This guide shows the manual method first, so you know exactly what you are measuring, and then what an AI visibility tracker such as Doclight automates.
Step 1: Write down the questions your buyers ask
Start with 10 to 20 questions. Take them from places where buyers already phrase their needs in their own words:
- questions your sales team hears in first calls
- "best tool for X" and "X vs Y" questions in your category
- problems your product solves, phrased as a question ("How do I find out why a page is not indexed?")
Write each question exactly once and keep the wording fixed. Small rewordings change the answer, so a list you edit every week cannot show a trend.
Step 2: Decide what counts
Before you ask anything, decide what you will record for each answer. Two things are worth separating:
- Mention: your brand name or domain appears in the answer.
- Citation: the answer links to your site, or names your domain as a source.
A brand can be mentioned without being cited, and the difference matters: a mention puts you in the buyer's head, a citation can also send them to your site. See what AI visibility actually measures for the longer version.
Also record which competitors are named, because "we were not mentioned" is much more useful next to "these three companies were".
Step 3: Ask, and record every answer
Use a fresh conversation for each question, with no custom instructions and no memory, so earlier chats do not shape the answer. For each question, record:
| Column | Example |
|---|---|
| Date | 2026-10-21 |
| Question | What are the best tools for tracking brand visibility in AI? |
| Mentioned? | No |
| Cited? | No |
| Brands named | Three competitors |
| Sources linked | The domains the answer cites |
Save the full answer text too. You will want it when a number looks strange.
Step 4: Repeat on a schedule
Run the same list again every week, on the same day if you can. After four to six runs you can see which questions you appear in reliably, which ones never, and which ones flip between runs. Report each result with its denominator: "mentioned in 6 of 20 answers", not "30%" on its own.
Step 5: Look at who is cited instead
For each question where you are missing, open the sources the answer links to. They are often comparison pages and "best tools" lists, not vendor homepages. Those pages are the ones worth getting onto, and they tell you what an answer engine found useful for that question. How to compare AI visibility with competitors covers how to make that comparison fair.
Where the manual method breaks down
The manual method is the right way to start, and it stops scaling at about the point it becomes useful:
- Twenty questions across four assistants is 80 answers per run, every week.
- Name matching by eye is inconsistent, especially for short or common brand names.
- ChatGPT is one assistant. Buyers also ask Claude, Perplexity and Gemini, and the answers differ.
- Comparing this month with last month means keeping the question wording and the counting rules identical across a growing spreadsheet.
What a tracker automates
Doclight is an AI visibility tracker. It sends your fixed question list to ChatGPT, Claude, Perplexity and Gemini through each provider's API on a schedule, stores every answer, and records mentions and citations separately for your brand and the competitors you name. When an answer is a numbered list, it records your position. Each answer is stored with the exact version of the question that produced it, so editing a question later does not rewrite history.
Two limits to keep in mind, whichever way you track:
- API answers are not the consumer app. Doclight asks through each provider's API. The ChatGPT app can answer differently, for example when it uses memory or search settings you do not control. See metrics and methodology.
- Name matching is exact. Misspellings and nicknames are not counted, so register the names people actually use.
Sources and verification
Product behavior described here was checked against the Doclight codebase on 2026-10-21. The example row in Step 3 is illustrative, not a real measurement.
Next: why your company might not show up in ChatGPT recommendations and a repeatable weekly monitoring routine.