How to check if ChatGPT mentions your brand (by hand and with an API)

To check if ChatGPT mentions your brand, ask it the questions your buyers ask, with web search on and in a temporary chat, and write down which brands it names and which pages it cites. Repeat each question three times, because answers change between runs, then do the same in Perplexity and Gemini. That takes about an hour for 10 questions. To repeat it every week without the hour, run the same questions through an API such as Cite42, which asks the models for you and returns the brands and sources as data for $0.13 per question.

This guide covers both: the manual check first, then the automated version, then what to do with the answer. Cite42 publishes it. The API examples are Cite42's own, and the manual steps need nothing but a browser.

Step 1: Write the questions buyers ask, not your brand name

Asking ChatGPT "what is Acme?" tells you whether it knows you exist. That is rarely the problem. The question that costs you customers is "what is the best invoicing software for freelancers?", answered with a shortlist that leaves you out.

Write 10 questions in the words a buyer would type:

• "Best [category] for [customer type]"
• "[Competitor] alternatives"
• "Cheapest [category] with [feature]"
• "[Category] with an API" or "[category] that works with [tool they already use]"
• "Is [competitor] worth it?"

Keep the list in a spreadsheet. You will reuse it every time you check.

Step 2: Ask ChatGPT in a temporary chat with search on

Open ChatGPT, start a temporary chat so your history and memory do not shape the answer, and turn on web search. Ask the first question exactly as written. Record three things: whether your brand appears, its position in the list, and the sources ChatGPT links. Then ask the same question again in a new temporary chat. Three runs per question is enough to see whether a mention is stable or a coin flip.

A few things to watch:

• Search on and search off are different products. Without search, ChatGPT answers from training data and tends to name the biggest brands. With search, it reads a handful of pages and names what those pages name. Check the version your buyers use, which in 2026 is mostly search on.
• Your account shapes the answer. Memory, custom instructions and past chats all nudge results, which is why the temporary chat matters. Signed-out results differ again.
• Record the sources. When a competitor is named and you are not, the cited page usually explains why: a comparison article, a directory listing, a Reddit thread. Those pages are your to-do list.

Step 3: Repeat in Perplexity, Gemini and Google AI Overviews

Perplexity shows its sources inline, which makes the record easy. Gemini and Google AI Mode draw on Google's index, so a brand that ranks well in Google often shows up there first. Google AI Overviews appear above the regular results for some queries and not others, so note when no overview appears at all.

Ask the same 10 questions in each engine and expect different answers from each one. One engine is not a proxy for the rest. Cite42's own baseline of 54 AI answers across three engines is written up in the tracking case study.

Step 4: Record the result as a table

Three numbers matter: your mention rate (answers that name you divided by all answers), your average position when named, and the list of pages cited in answers that leave you out. The first two tell you where you stand, and the third tells you what to do.

Step 5: Automate the check so you can repeat it weekly

The manual check is free and worth doing once. It does not scale: 10 questions, three engines and three runs each is 90 prompts, and the answers you wrote down last month are already stale.

Cite42 runs the same check as one API call per question. You send the question and the brands to look for, pick the engines, and get back the mention rate, average position and per-model rank for each brand. A check across ChatGPT, Perplexity and Gemini costs $0.13. All five engines, including Claude and Google AI Overviews, cost $0.20. There is no subscription, you pay only for successful calls, and a verified signup includes $1.00 of credit.

From Claude, ChatGPT, Codex or Cursor. Connect the Cite42 MCP server (hosted at https://mcp.cite42.dev, sign in with your account) and ask in plain English:

Does ChatGPT mention Acme for "best invoicing software for freelancers"? Compare with FreshBooks and Wave.

The assistant calls cite42_rankings and gives you the table. Ask which pages the answers cited and it calls cite42_citations. The setup docs list the one-line install for each app.

From your own code. The same tool is a REST endpoint:

curl -X POST https://www.cite42.dev/api/v1/rankings \
  -H "Authorization: Bearer cite42_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "best invoicing software for freelancers",
    "competitors": ["Acme", "FreshBooks", "Wave"],
    "models": ["chatgpt", "perplexity", "gemini"]
  }'

The response lists each brand with its mentionRate, avgPosition and byModel rank, so a short script can fill the table from Step 4 for all 10 questions in a minute.

On a schedule. Save the 10 questions as an AI Visibility tracker and it reruns them weekly or monthly and keeps the history per engine. Ten questions a week across three engines averages $5.63 a month.

If you would rather have a dashboard, OtterlyAI, the Semrush AI Visibility Toolkit and Ahrefs Brand Radar track prompts daily on monthly plans from $29 to $99 (as of October 2026). The tool comparison and the pricing guide cover them.

Step 6: Turn each missing mention into a page

Checking more often does not raise a low mention rate. What raises it is appearing on the pages the engines cite. For each question where you are absent, open the cited sources and ask: can we be listed here (a directory, a comparison, a Reddit thread), or do we need our own page that answers this question plainly? The content gap workflow walks through that with keyword demand added in.

Then check again in two to four weeks. Engines with search pick up new pages within days, but a mention usually needs a few weeks of being cited before it becomes stable.

Frequently asked questions

How do I know if ChatGPT recommends my brand?

Ask ChatGPT the buying questions your customers ask, in a temporary chat with web search on, and note whether your brand is named and in what position. Repeat each question three times, because answers vary. Cite42 automates the same check and returns the mention rate and rank per AI model, for $0.05 per question on ChatGPT alone.

Can I check AI brand mentions for free?

Yes. The manual method in this guide costs nothing but time. Cite42 adds $1.00 of credit to a verified account, which covers 7 questions across ChatGPT, Perplexity and Gemini, with no subscription afterwards. OtterlyAI, Peec AI, Profound and Scrunch offer free trials of their dashboards.

Why does ChatGPT give different answers to the same question?

Because each answer is generated fresh and, with search on, from whichever pages the search step returned that time. Personalization, memory and A/B tests add more variation in the consumer app. Treat a single answer as a sample and measure a mention rate across several runs rather than a yes or no.

How often should I check whether AI assistants mention my brand?

Weekly is enough for most companies, and monthly if your category moves slowly. Check right after publishing a page that targets one of your questions, and again two to four weeks later. Daily tracking mainly matters in news-driven categories.

Does the ChatGPT API give the same answers as the ChatGPT app?

Not exactly. Cite42 asks each model through its official API with web search on, which is the closest repeatable version of what people see. The app adds personalization, memory and experiments that an API does not reproduce, so use API results as a consistent measurement rather than a copy of one person's chat.

Run your first 10 questions

Write the questions today and run them by hand once. Then create a free Cite42 account for $1.00 of credit, connect the MCP server to Claude, ChatGPT, Codex or Cursor, and run the same list in one go. Save it as a weekly tracker once you have a page to measure.