How to track whether ChatGPT and Gemini mention your brand
You can’t see AI answers in a rank tracker, and no assistant reports how often it names you. But with a fixed set of prompts, a simple log and a few analytics settings, you can measure it well enough to act on.
ARAgile Ranker TeamSearch & AI Visibility deskPublished9 Oct 2026Read8 minReviewed bySearch & AI Visibility desk
One prompt, five assistants — this month
ChatGPT: named second, described accurately
Gemini: named, but shows the old address
Google AI Mode: named, own page cited
Perplexity: not named, competitor cited
Copilot: named, review site cited
Skim · 2 minFull · 8 minShare
The quick answerCopy
Build a fixed set of 20 to 50 buying prompts, run them on each assistant on a regular schedule, and log whether you are named, how you are described and which sources are cited. Pair that with AI referral traffic in GA4. Judge trends across many prompts — single answers vary too much to mean anything.
Track buying prompts — the questions people ask before contacting anyone — not just your brand name.
Answers vary between runs, users and locations, so report patterns across many prompts, never one screenshot.
Log how you are described and which sources are cited; that shows what the model believes and where it learned it.
AI referrals show up in GA4 under the assistants’ domains, but many AI-influenced visits arrive as direct or branded search.
Start manually for a month, then move to a tool once you know which prompts and engines matter.
01
Why track AI mentions separately from SEO rankings?
Because an assistant can recommend a competitor while you still rank first on Google. AI answers draw on different signals — third-party mentions, consistent facts, crawler access — so rankings alone will not tell you whether you are on the shortlist.
A ranking report answers “where does my page appear for this keyword?” An AI answer is a different product: a short written recommendation, assembled from several sources, often naming two to five businesses. Your page can be in the sources and your brand still left out of the text — or the reverse.
Buyers increasingly ask an assistant for a shortlist before they search for individual businesses. If you are missing from that shortlist, you will see it later as fewer branded searches and fewer enquiries, with no obvious cause in your SEO reports. Tracking mentions directly is the only way to see the problem while it is still small.
02
Which prompts should you track?
The questions a buyer asks before contacting anyone: “best [service] in [city]”, “[service] for [situation]”, comparisons, and cost questions. Write 20 to 50 in your buyers’ own words, grouped by intent, and keep them fixed so results are comparable month to month.
Start from real language. A buyer in Jaipur might ask “Which CA firm is good for startups in Jaipur that also handles GST?”, not “chartered accountant Jaipur”. Prompts are longer and more conversational than keywords, and the extra context — budget, location, situation — changes which brands appear.
Group prompts so you can see where you are strong and weak. Add a few brand prompts too, such as “What do you know about [your brand]?” and “[your brand] reviews”, to check the facts the assistant holds about you. Then freeze the set. Changing prompts every month makes trends meaningless; add new ones in a separate group instead.
Category: “best [service] in [city]”
Situation: “[service] for [type of customer or problem]”
Comparison: “[you] vs [alternative approach]” or “X or Y for…”
Cost: “how much does [service] cost in [city]”
Brand facts: “what do you know about [brand]?”
03
Why does ChatGPT give a different answer every time?
Assistants generate text probabilistically, and search-enabled answers depend on which pages are retrieved that day, plus location, language and the user’s history. One answer proves little; a pattern across many prompts and repeated runs is what to report.
Ask the same question three times and you may get three different lists. That is normal, not a bug. It means a single screenshot — good or bad — should never drive a decision, and it means your tracking method needs repetition built in.
Reduce the noise you can control. Use a temporary or logged-out chat so personal memory and history do not shape the answer. Set the location the way your buyers would experience it, or include the city in the prompt. Run each prompt more than once per check — three runs is a sensible minimum — and record how many of those runs named you. Over a month, that gives you a mention rate per prompt rather than a yes or no.
Want this done for you?
Our AI Search Optimisation team can take it on — or start with the free AI visibility check and see where you stand on Google and five AI engines. No obligation.
Whether you are named, your position in the list, how you are described, which competitors appear, and which sources are cited. Descriptions and sources matter as much as mentions — they show what the model believes and where it learned it.
A simple spreadsheet works: one row per prompt per engine per run, with columns for date, prompt group, engine, named (yes or no), position, description summary, competitors named, and sources cited. After a month, pivot it by engine and by prompt group.
The source column is where the action is. If Gemini keeps citing a directory with your old address, that is a fix you can make this week. If every engine cites the same two comparison articles and you are not in either, you know which publishers to approach. If the description of your business is wrong or vague, your own pages are probably not stating it plainly enough.
05
Should you track manually or use a tool?
Start manually with a spreadsheet for a month to learn which prompts and engines matter. Move to a tracking tool once you have more than about 50 prompts or several markets — but understand how a tool samples before trusting its scores.
Manual checks are slow, but they teach you things a dashboard hides: how the answers are phrased, which sources keep appearing, where the facts about you are wrong. A month of manual tracking also gives you a baseline to check a tool against.
AI visibility tools run prompts at scale, across engines and locations, and chart a share-of-voice figure over time. Ask how often they run each prompt, from which locations, whether they use the consumer apps or the developer APIs (answers can differ), and whether you can see the raw answers behind the score. A score you cannot trace back to real answers is hard to act on.
06
Can you see AI assistant traffic in Google Analytics?
Partly. Visits from links in ChatGPT, Perplexity, Gemini, Copilot and Claude show as referrals from their domains, and ChatGPT often tags links with utm_source=chatgpt.com. Many AI-influenced visits arrive as direct or branded search, so treat the referral number as a floor.
In GA4, create a custom channel group with an “AI assistants” channel that matches session sources such as chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai. Place it above Referral in the channel order so these sessions are not lumped in with other sites. Then mark your key events — enquiries, bookings, purchases — so you can see whether this traffic converts.
Google’s own AI Overviews and AI Mode are different: clicks from them arrive as ordinary Google organic traffic, and Search Console includes their impressions in its Web search totals without a separate filter. Watch branded search volume alongside AI referrals; a rise in people searching your name is often the clearest downstream sign that assistants are mentioning you.
07
How do you set up a monthly AI visibility check?
1Write and freeze your prompt setDraft 20 to 50 prompts from real customer questions, group them by intent, and save them in a sheet. Keep the wording fixed from month to month.
2Choose the engines that matterUsually ChatGPT and Google’s AI features first, then Gemini, Perplexity, Copilot and Claude, depending on your buyers and market.
3Run each prompt three times per engineUse a temporary or logged-out session, note the location, and record every run — not just the best or worst one.
4Log mentions, descriptions and sourcesRecord named or not, position, how you were described, competitors named and cited sources for every run.
5Add the analytics viewSet up an AI assistants channel in GA4 and a monthly note of branded search impressions from Search Console.
6Review trends and assign fixesEach month, compare mention rates by engine and prompt group, then turn wrong facts and missing sources into a short fix list.
08
Questions people ask
How often should I check AI mentions?
Monthly is enough for most businesses. Answers shift as models update and new pages are indexed, but week-to-week changes are mostly noise. Check more often only after a big change, such as a rebrand, a site migration or a major PR win.
Does being logged in change what ChatGPT says?
It can. Memory, past conversations and custom instructions can shape answers for a logged-in user. For tracking, use a temporary chat or a logged-out session so results reflect what a new buyer would see, not what the assistant has learned about you.
Is there an official report showing how often ChatGPT mentions my brand?
No. None of the major assistants currently offers businesses a report of how often they are mentioned in answers. That is why tracking relies on running prompts yourself or through a tool, plus referral data in your analytics.
Does Search Console show AI Overview performance separately?
Not at the time of writing. Impressions and clicks from AI Overviews and AI Mode are counted within Search Console’s Web search totals. You can watch the queries that trigger AI features, but you cannot filter them out as a separate report.
What is a good mention rate to aim for?
There is no universal benchmark; it depends on how crowded your category is and how specific the prompts are. Measure your own baseline, compare it with the competitors that appear most, and aim to improve prompt group by prompt group.
Cite this article: Agile Ranker Team (2026). How to track whether ChatGPT and Gemini mention your brand. Agile Ranker Insights. Published 9 Oct 2026.
AR
Written by
Agile Ranker TeamSearch & AI Visibility desk · Agile Ranker
The Agile Ranker team that runs search, AI visibility, paid media and content for clients in ten markets. Notes are checked against current platform documentation before publishing.
Reviewed by the Search & AI Visibility desk
Included with every plan
A free 5-page website for your business.
A fast, AI-readable site that turns the traffic we bring into enquiries — built to the same standard as our paid web work.
5 pages, your choiceHome, About, Services, Key service, Contact
Free with any monthly plan from $499. Domain, hosting and paid tools are bought by you and stay in your name; shops and booking systems are quoted separately. The 48-hour clock starts once your plan is active and the launch checklist is done.
yourbusiness.comLIVE
HomeAboutServicesKey serviceContact
Your logo
Book a call
Schemallms.txtGA4Fast on mobile
FREEwith any plan
48h
Keep reading
Three notes that pick up from here.
Each one opens with its quick answer — skim first, read if it's yours.