ChatGPT Visibility Tracker: How to Monitor Your Brand

The best way to track whether ChatGPT recommends your brand is a ChatGPT visibility tracker: a fixed set of buyer prompts, run repeatedly on a schedule, with mentions, cited sources and competitors recorded every time. A single manual check misleads, because ChatGPT gives a different answer almost every time you ask.

Key takeaways

• Track four things: mention rate, citation rate, the source pages behind answers, and prompt variance. A tracker that shows only one of them leaves you guessing.
• Never judge from one run. Repeat each prompt several times and report a rate with a margin of error, not a yes or no.
• Use 20 to 50 unbranded buyer prompts, wording frozen, so a change in the numbers is a change in ChatGPT and not in your questions.
• Run daily if you can, weekly at minimum, and read the trend over 7 days or more rather than any single day.
• Treat "ranking position in ChatGPT" with caution: SparkToro's research found the order of brand lists changes from run to run.
• Start with a free check, then decide whether you need a daily tracker. Noma's free brand audit shows how AI engines describe your brand in minutes.

What does a ChatGPT visibility tracker actually measure?

A ChatGPT visibility tracker asks ChatGPT the questions your buyers ask, stores each answer, and turns the stored answers into rates you can compare over time. It is not a rank tracker. There is no results page with ten blue links and a fixed position one. There is an answer, and your brand is either inside it or not.

Four measures matter. They sit on top of the definitions in our guide to what AI visibility is.

Mentions and citations are separate outcomes. A brand can be named with no link, and a page can be cited while a competitor gets the recommendation. Citation vs brand mention explains why each calls for different work. Add share of voice, your mentions as a share of all brand mentions in the prompt set, once you track competitors: how AI share of voice is calculated has the formula.

Why does one manual check of ChatGPT mislead?

One check misleads because ChatGPT is not a lookup. The same prompt returns a different brand list, a different order and a different length on different runs. If you saw your brand once, you may have been lucky. If you did not see it, you may have been unlucky.

SparkToro and Gumshoe.ai tested this. As Search Engine Journal, January 2026 reports, they ran 2,961 prompts across ChatGPT, Claude and Google's AI features, running each of 12 brand-recommendation prompts 60 to 100 times per platform. Nearly every response was unique in the brands listed, their order and the number of items.

The same research found something more useful: how often a brand appears across many runs is steadier than any one answer. For varied headphone prompts, Bose, Sony, Sennheiser and Apple appeared in 55 to 77% of 994 responses, according to the same report. That is the case for tracking a rate, not a result.

Rand Fishkin, SparkToro co-founder, as quoted by Search Engine Journal: "any tool that gives a 'ranking position in AI' is full of baloney."

Citations move too. In a 20-day study of 37,500 citations, Ahrefs, October 2026 found that around 60% of citations on ChatGPT and Google AI Mode were gone by the next day. It also found that about 75% of the disappeared citations came back within 20 days. Its advice: do not judge AI visibility from one check, and count how many of the last seven days you were cited.

A page that is cited today and missing tomorrow has not necessarily been dropped. A page edit followed by one comeback is not proof the edit worked. Our note on AI citation stability covers how to separate noise from change.

How do I build a ChatGPT visibility tracker?

You can build a working tracker with a spreadsheet and one hour a week, or run it in a tool. The method is the same either way. These six steps follow the approach in the AI Visibility Measurement Standard.

Step 1: Write 20 to 50 unbranded buyer prompts

Use the questions a buyer asks before they know your name: "best project management software for a 20-person agency", not "is [Your Brand] good". Branded prompts tell you how ChatGPT describes you. Unbranded prompts tell you whether you get on the shortlist. Tag each prompt by intent: category discovery, comparison, problem-led, and pricing. Which prompts to track goes deeper, and the free AI Prompt Finder suggests buyer questions for your category, tagged by intent.

Step 2: Freeze the wording

Do not tweak prompts after the baseline. If you change the words, you have changed the instrument. SparkToro's volunteers wrote headphone prompts with a semantic similarity score of only 0.081, per the Search Engine Journal report, yet ChatGPT-style tools still returned a similar consideration set. So buyers will phrase things differently from you. Cover that by adding two or three paraphrases of your most important prompts, each tracked separately.

Step 3: Run each prompt several times, from a clean session

Use a fresh chat each time, logged out or with memory and custom instructions off, so earlier conversation does not shape the answer. Run each prompt at least 5 times per measurement period. Record whether web search ran and show the cited links, because ChatGPT answers with and without live search draw on different material.

Step 4: Log the same fields for every answer

• Date, prompt ID, run number
• Brands named, in order of appearance
• Whether your brand was named (yes or no)
• URLs cited, and whether any are yours
• The words used about your brand (positive, neutral, wrong)

Step 5: Calculate rates and a margin of error

Worked example, with illustrative numbers: 20 prompts, 5 runs each, gives 100 answers. Your brand appears in 18. Mention rate is 18%. A rough 95% margin of error is 1.96 × the square root of (0.18 × 0.82 / 100), about 7.5 points, so the true rate is plausibly 10 to 26%. Repeat runs of one prompt are not fully independent, so treat that as a floor on the uncertainty. A move from 18% to 22% next week is inside the noise. A move to 35% is not.

Step 6: Review the sources, not just the score

List every domain cited across the answers and rank by frequency. If review sites, comparison articles or forum threads dominate, your work is on other people's pages. How AI engines decide which sources to cite explains the logic.

How often should I check ChatGPT visibility?

Check as often as you can afford to keep the prompt set and wording identical. Daily runs give you the 7-day windows Ahrefs recommends. Weekly is the floor. The table gives a practical split.

After a model release, expect movement that has nothing to do with your work. What a new AI model release changes for your brand's visibility covers the first-week checks.

What should a ChatGPT tracker show beyond a mention count?

A useful tracker shows why the number moved, not only that it did. Use this checklist when you compare tools or build your own.

• Full answer text saved. You need to read what ChatGPT said, not only a score.
• Cited URLs per answer. Needed for the source mix.
• Repeat runs. One run per prompt per day is a sample of one.
• Competitor mentions. Your rate means little without theirs.
• Sentiment or description accuracy. Being named wrongly can cost you more than being absent. See how to fix negative brand sentiment in AI answers.
• Other engines. Buyers do not use only one assistant, and engines disagree. Why ChatGPT, Gemini, Claude and Perplexity disagree about your brand explains the gap.

Vendors describe this differently. As checked on 2026-11-03, SE Ranking's ChatGPT Visibility Tracker page lists detecting brand mentions and links, seeing which sources ChatGPT relies on, and seeing where a brand ranks in ChatGPT answers. Otterly publishes a guide to brand monitoring on ChatGPT. If a tool reports a single "position", ask how many runs sit behind it, given the variance above.

Does being crawlable by OpenAI change what you can track?

Yes, because a tracker shows the result and crawl access is one of the inputs. OpenAI documents its crawlers on its Overview of OpenAI Crawlers page, and the user agents for training, search and user-initiated fetches are separate. A firewall or CDN rule that blocks the wrong one can quietly remove you from ChatGPT's search-based answers while your robots.txt looks fine.

Before you read any visibility number, run the free AI Crawler Checker. It shows whether GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot can reach your site and generates a corrected robots.txt. If ChatGPT cannot fetch a page, that page is unlikely to be cited. For symptoms and fixes, see ChatGPT agent blocked your site?

How do I run a free ChatGPT visibility check today?

Do this in 30 minutes before paying for anything.

1. Open the free brand audit to see how AI engines describe your brand and who they name instead.
2. Write 10 unbranded buyer prompts and run each 3 times in fresh ChatGPT chats.
3. Fill in a sheet with the fields from step 4 above.
4. Note your mention rate, the top five cited domains and every competitor named.
5. Repeat a week later with identical wording. If the rate moved by less than your margin of error, you have a baseline, not a trend.

If the spreadsheet already takes more than an hour a week, that is the signal to automate. How to track brand mentions in ChatGPT has the full manual method, including chatgpt.com referrals in GA4.

Which approach fits: manual, single-engine tool or multi-engine platform?

Most buyer journeys now touch more than one assistant, so a ChatGPT-only view can hide a problem in Gemini or Perplexity. How to track your brand across AI engines covers the wider set.

Where does Noma fit?

Noma by Nurdd is an answer engine optimisation platform that asks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, the questions your buyers ask, every day. For ChatGPT it reports visibility, position and sentiment, names the competitors that appear instead of you, and lists the sources each answer cites. ChatGPT is tracked on every paid plan. See the ChatGPT visibility tracker page for detail.

See it on your own brand: book a demo or start the 3-day free trial. A card is taken at signup so billing can start on day 3; cancel before then and nothing is charged. Plans are on the pricing page.

What mistakes make ChatGPT tracking useless?

• Tracking branded prompts only. You will always look visible.
• Rewording prompts mid-way. You lose comparability.
• Reporting position as a fact. Order changes run to run.
• Reacting to one day. Citations drop and return.
• Ignoring wrong answers. Accuracy is a measure, not a footnote. The sequence for correcting wrong AI answers starts with the source the engine relied on.
• Not connecting to action. Each finding needs an owner: content, PR, web or product marketing. How to rank in ChatGPT lists the fixes.

How we researched this

We reviewed OpenAI's published crawler documentation, a Search Engine Journal report on SparkToro and Gumshoe.ai's research, Ahrefs' citation stability study, and the public ChatGPT tracking pages and guides of SE Ranking and Otterly. Statistics are quoted as the sources state them, with the publication month. Vendor facts and Noma's plan details were checked on 2026-11-03. The worked example is illustrative arithmetic, not measured data.

What to do next

Run the free brand audit, write your 20 prompts, and take a one-week baseline with repeat runs. Then decide whether to automate it. To see daily tracking across 13 AI engines on your own brand, book a demo or start the free trial.