How to Track Your Brand in ChatGPT, Perplexity, Gemini and Claude

To track brand mentions in ChatGPT, Perplexity, Gemini and Claude, pick a fixed set of customer questions, ask each one in every engine on a regular schedule, and record whether you were mentioned, where you appeared, which competitors were named and which sources were cited. A simple spreadsheet is enough to start; automated tracking helps once the list grows.

Below is a step-by-step manual method, what to record, how often to check, and the mistakes that make results misleading.

Why track brand visibility in AI search at all?

More people now ask AI assistants for recommendations instead of scrolling through search results. The assistant names a handful of businesses and moves on. If you are not in that short answer, the person may never see you.

Your website analytics will not show you this. There is no impression report for “ChatGPT considered you and chose someone else”. The only way to know your brand visibility in AI search is to ask the questions yourself and write down what comes back.

How do you track brand mentions in ChatGPT manually?

Step 1: Choose 10 to 20 prompts

Use real customer questions, not your brand name. Mix intents:

• Local: “Best family dentist in Austin?”
• Ready-to-buy: “Which bookkeeping service should a small café in London use?”
• Comparison: “Is [competitor] or [you] better for small teams?”
• Brand: “What is [your business]?” and “How can I contact [your business]?”

If you are not sure where to find good prompts, our guide to AI prompt research walks through sales calls, reviews, FAQs and competitor names.

Step 2: Set up a simple spreadsheet

One row per answer. These columns are enough:

Step 3: Ask each prompt in each engine

Open ChatGPT, Perplexity, Gemini and Claude. Where the engine offers it, make sure web search is on, because that is when it retrieves pages and cites sources. Paste the prompt exactly as written in your sheet. Do not rephrase it between engines or between weeks.

Step 4: Repeat each prompt more than once

Ask the same prompt two or three times, ideally in fresh chats. If you appear in one answer out of three, that is very different from appearing every time. A single check can make you look better or worse than you really are.

Step 5: Record everything, including the sources

Fill in a row for every answer. Copy the cited links, not just the brand names. The sources tell you why the engine said what it said.

Step 6: Summarise per prompt

After each round, count how often you were mentioned out of the answers you collected for each prompt and engine. Highlight gaps: answers where a competitor was named and you were not.

Why do AI results vary so much?

If you check the same question twice and get different answers, nothing is broken. AI answers vary by:

• Engine: each uses its own models, search index and sources.
• Wording: “best” and “most affordable” can produce different lists.
• Location: local questions depend on where the engine thinks you are.
• Time: the web changes, and so do the pages the engine retrieves.
• Personalisation: memory, chat history and account settings can shape answers.
• Access method: answers through an API can differ from the consumer app.

No AI company publishes its ranking formula, so treat any single answer as a sample, not a verdict. That is why repeated checks and fixed prompts matter.

What should you look for in each answer?

Were you mentioned?

The basic yes or no. Track it per prompt and per engine, because you may be strong in Perplexity and missing in Gemini.

What position were you in?

Being first in a list of five is not the same as being a footnote. Position is rough, but useful over time.

Which competitors appeared?

Note who is named when you are not. Over a few rounds you will see which competitors AI engines lean on, which is the starting point for measuring AI share of voice.

Which sources were cited?

Cited pages are often directories, “best of” lists, review sites, news articles or forum threads. If those pages mention competitors and not you, that is a concrete gap you can work on. See which websites ChatGPT cites for the common types.

Are your facts right?

For brand prompts, check the phone number, website, email, address and founding year. Wrong details are common when your information is inconsistent across the web. Our guide on fixing wrong information in ChatGPT explains the next steps.

How often should you monitor ChatGPT mentions?

• Monthly works for most small businesses with a stable market.
• Weekly suits competitive categories, active campaigns, or when you have just made changes and want to see whether they help.
• After major changes, such as a rebrand, new location or new website, run an extra round.

Whatever you choose, keep it consistent. Same prompts, same engines, similar settings. Our guide to measuring AI visibility improvements covers how to read trends without over-reacting.

What are the common pitfalls?

• Checking once and drawing conclusions. One answer is a snapshot. Look at several answers over several rounds.
• Testing while logged in to your own account. Your history may include your business, which can tilt answers. Use a fresh or logged-out session where possible.
• Only asking branded questions. “What is [your business]?” tests accuracy, not whether you get recommended. Most prompts should be unbranded.
• Changing the wording each time. New wording means a new question, and you lose your trend.
• Ignoring the sources. The cited pages are where most of your next actions come from.
• Expecting guarantees. No one can promise a mention. You can improve your odds and measure the result.

When should you switch to automated AI brand monitoring?

The manual method is a good way to learn, but it gets heavy quickly. Twenty prompts across four engines, asked three times each, is already a lot of copying and pasting every round. Automated AI brand monitoring makes sense when:

• you track more than a handful of prompts or several locations,
• you need the same schedule every week without someone remembering to do it,
• you want cited sources and competitor names stored automatically, or
• you manage several brands, as agencies do.

Look for tools that use real answers from the engines, keep the full answer and sources, and do not show a rate from a single answer.

How Onxeera helps

Onxeera’s Prompt Research asks ChatGPT, Perplexity, Gemini and Claude through their official APIs with web search on, and stores each answer, the brands mentioned, their order and the cited sources. A mention rate appears only after 3 answers, and paid plans track prompts automatically. For brand questions, AI Reputation flags wrong phone numbers, websites, emails or founding years with the quote and sources. There is a free plan to start.

Frequently asked questions

Can I use Google Alerts to track ChatGPT mentions?

No. Google Alerts watches web pages, not AI answers. AI assistants generate answers on demand, so the only way to see them is to ask the questions and record the replies, either by hand or with a tool that queries the engines for you.

Do ChatGPT and Perplexity give everyone the same answer?

No. Answers can vary by wording, location, time, personalisation and the engine's own sources. Two people asking the same question may see different businesses. That is why you should run each prompt more than once and look at patterns rather than single answers.

How many engines should I track?

Start with ChatGPT, Perplexity, Gemini and Claude, since they are widely used and can search the web. If your customers rely heavily on Google, also check AI Overviews and AI Mode, and consider Microsoft Copilot if your audience uses Bing or Windows a lot.

What should I do if I am never mentioned?

Check that AI search crawlers can access your site, make your business facts clear and consistent, and look at the sources cited for your prompts. Getting listed or mentioned on those pages often matters as much as changes to your own website.