Citino teamTo track your brand’s visibility in ChatGPT, ask it the same buyer questions on a schedule and record what each answer says about you. One answer tells you little, because ChatGPT’s answers change from run to run. A rate over many answers, with a margin of error, tells you where you stand and whether it is changing.
This is one engine’s part of AI search analytics. The method is the same for Perplexity and Google AI Overviews; the details below are specific to ChatGPT.
What brand visibility in ChatGPT means
Four numbers describe how ChatGPT treats your brand, counted over the answers to the questions you track:
| Number | The question it answers |
|---|---|
| Visibility | What share of answers name you at all? |
| Share of voice | What share of all brand mentions are yours? |
| Position | How early in the answer are you named? |
| Sentiment | How does the answer describe you? |
Behind them sit the pages ChatGPT cites, which tell you why it answered the way it did.
Step 1: Choose the questions to track
Track the questions a buyer asks ChatGPT while choosing a product like yours, in their words: category questions (“best WMS for a 3PL”), problem questions, comparisons, alternatives and fact checks. Keep branded questions in a separate group, because a question that names you will almost always get an answer that names you.
For one brand, 25 to 150 questions covers most buying decisions. How to choose the prompts to track explains how to find and word them.
Step 2: Ask the same way every time
- Fix the wording. Language models are sensitive to small changes in a prompt (Sclar et al.), so a reworded question is a different measurement. Start a new series rather than editing an old one.
- Use a clean session. Saved memory, custom instructions and earlier messages in a chat can all shape an answer. Ask each question in a new chat, with nothing saved, so you measure what a new buyer would see.
- Keep the time fixed. Run the questions at the same time each day, so a change in the result is not a change in when you asked.
Step 3: Note whether ChatGPT searched the web
ChatGPT answers some questions from what the model learned in training and searches the web for others. The two behave differently:
| Answer from search | Answer from training data | |
|---|---|---|
| Sources | Cited links you can open | None |
| Changes when | The pages it reads change | A new model ships |
| What you can do | Fix or get onto the cited pages | Make sure future training sources are right |
OpenAI’s documentation describes how its models use web search to answer with sourced citations, each citation carrying the URL and title of the page. Record those pages for every answer that has them: they are the most useful thing in the data, because they show which sites decide the answer.
Step 4: Record every answer
For each question and each run, keep:
| Field | Example (illustrative) |
|---|---|
| Date and time | 2026-10-23, 09:00 UTC |
| Question | “best warehouse management software for a 3PL” |
| Full answer text | The whole response, not a summary |
| Brands named, in order | Crateline, Northport, Tallyfold |
| Your position | 3rd |
| How you were described | “a simpler option for smaller 3PLs” |
| Searched the web? | Yes |
| Pages cited | The URLs in the answer |
Keep the full text. A number you cannot trace back to the answers behind it is hard to trust, and the wording is often where the useful detail is.
Step 5: Count, with a margin of error
Your visibility is the share of answers that name you. Report it with a margin of error, and calculate that margin over questions rather than over answers: answers to the same question move together, so treating each one as independent makes the result look more precise than it is. How to measure AI visibility works through the method.
A change smaller than its margin is not a change. If your visibility moves by less than the margin, label it within noise and do not report it as a win or a loss.
Manual tracking, scripts or a tool
| Approach | Works for | Watch out for |
|---|---|---|
| By hand, in a spreadsheet | Up to about ten questions, checked weekly | Slow, easy to drift from a fixed routine |
| A script against the API | Teams with engineering time | The API is a separate product from the ChatGPT app; check its answers against what you see in the app |
| An AI search analytics tool | Daily tracking across engines | Ask how it collects answers and how it calculates the margin of error |
Make sure ChatGPT can read you
Tracking only helps if ChatGPT can reach your pages. OpenAI says it uses OAI-SearchBot to surface websites in ChatGPT’s search features, and that sites which opt out of it will not be shown in ChatGPT search answers. Check your robots.txt allows it. How to get your brand recommended by ChatGPT covers what to change once you know where you stand.
Where Citino helps
Citino is an AI search analytics tool. It asks ChatGPT, Google AI Overviews and Perplexity your buyers’ questions every day, with Gemini, Google AI Mode and Claude as add-ons, from a clean session with fixed wording, and keeps every answer with the brands it named and the pages it cited. Each number carries a 95% margin of error bootstrapped over prompts, as the methodology sets out. See how Citino measures.