How to track Google AI Overviews
What to measure, why one check tells you nothing, and how to tell being cited apart from being named. The four states, and what each one means.
AI mention tracker platform
Track your AI mentions and find out how to get mentioned more
When someone asks ChatGPT, Claude, Gemini or Perplexity about your category, do you come up? We run those questions continuously, record what each assistant answered and cited, and tell you which of the three problems is keeping you out, so you know what to fix.
Card required, nothing charged for 7 days
- Six engines, never averaged
- Every rate with its denominator
- Answers stored verbatim
Overview · last 30 days · 1,240 answers
| Engine | Named | Cited | n |
|---|---|---|---|
| ChatGPT | 47% | 63% | 30 |
| Claude | 30% | 37% | 30 |
| Gemini | 57% | 70% | 30 |
| Perplexity | 40% | 80% | 30 |
| Google AI Overviews | 20% | 50% | 30 |
| Google AI Mode | 3/4 | 4/4 | 4 |
Google AI Mode shows a fraction: below five answers a percentage would be a guess.
Paste your domain. We propose the rest.
No tag to install, no analytics to connect, no call to book. We read your site and come back with the category a buyer would use, up to six competitors you genuinely lose deals to, and eight to twelve questions a buyer asks before they know your name — each with its real monthly search volume beside it. Every suggestion is a suggestion; nothing is tracked until you say so.
- 1
Give us the domain
That is the whole of the setup you cannot skip.
- 2
We read your site
And describe your category in a buyer's words, not your marketing's.
- 3
We propose the set
Competitors and buyer questions, priced by how many people actually ask them.
- 4
You edit and confirm
Change anything, then the first answers arrive within minutes.
Per-engine visibility
Six engines, reported separately
ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode retrieve differently and disagree with each other. A blended score hides which one you are losing. See how each platform is tracked.
- Never averaged.
- Each engine gets its own rate, its own trend and its own sample size. There is no single score, on purpose.
- Collection method recorded.
- A scraped product surface and a model API answer are different measurements. Every number says which it is.
- On a fixed schedule.
- The same questions, six days a week, so a change in the answer means something changed in the world rather than in the wording.
Engines · named · last 30 days
ChatGPT
47%n=30
Search product surface
Claude
30%n=30
API with web search
Gemini
57%n=30
Search product surface
Perplexity
40%n=30
API with web search
Google AI Overviews
20%n=30
Live SERP
Google AI Mode
3/4n=4
Live SERP
Answers, not scores
Every answer stored verbatim, with its citations
We keep the whole response: which brands were named, in what order, and which URLs the assistant actually read to get there. Read how a run works end to end.
- Full text.
- The answer as the assistant gave it, so you can see the framing rather than guess it from a number.
- Brands in order.
- First-named and last-named are different outcomes. We keep the position of every brand in every answer.
- Sources cited.
- Every URL the assistant cited, so you can see whose pages are shaping the answer in your category.
Answer · Perplexity · “best project tracker for agencies”
For agency work the strongest options are Northstar, Planwise and Tandem. Northstar handles multi-client billing best, while Planwise has the clearest approval flow for client sign-off…
Sources cited
- northstar.com/agenciesnamed
- yourbrand.com/blog/agency-billingcited, not named
- planwise.io/approvalsnamed
- g2.com/categories/project-trackingthird party
One workspace, every assistant
Six surfaces, and we record how each answer was collected. See all platforms or the measurement methodology.
Cited but not named
We tell you which problem you have
Not being mentioned is three different problems with three different fixes. An assistant can read your page, use what it learned, and recommend a competitor by name. A blended score reports that as a low number. We report it as its own state, because the fix is completely different from the fix for never being retrieved at all.
- Never retrieved.
- The assistant did not see your page. A discoverability problem: crawlability, indexing, or a topic you have not written about.
- Retrieved, not cited.
- Fetched and discarded. Your page was a candidate and lost to a better one.
- Cited, not named.
- Your content was good enough to use and not framed in a way the model repeats. The fix is on the page, not in the index.
States · ChatGPT · 30 answers
- 14 / 30
Named
You were recommended by name.
- 7 / 30
Cited, not named
Your page was read. Someone else got the credit.
- 4 / 30
Retrieved, not cited
Fetched, then not used.
- 5 / 30
Not retrieved
The assistant never saw you.
Honest numbers
Every rate carries its sample size
One check of an AI answer is a draw, not a rate. Every figure on every screen shows the number of answers behind it. The rules are published in full in the measurement methodology.
- Fractions below five.
- Under five answers we show 3/4, not 75%. A percentage there would be a guess dressed as a measurement.
- Failures excluded.
- A collection error on our side is dropped from the denominator, never counted as your brand being absent.
- Alerts on real change.
- Email when a rate moves beyond what the sample can explain, not on every wobble.
Sample · “which CRM should a small law firm use”
One check
1 / 1
Shown as a fraction. A single answer is a draw, not a rate.
14 recent runs
71%
10 of 14 answers named you. Now it is a rate.
Same question, same engine, one answer per day. Failures on our side are excluded from the denominator.
Google AI Overviews
Track the answer above the results
A reader can finish a Google search without scrolling to a single link. Whether you were part of the AI Overview is now a separate question from where you rank. See the AI Overview tracker and how Overviews are collected.
- Live SERP.
- Collected from the real result page, so it moves when Google moves, not when a model is retrained.
- AI Mode too.
- Google's conversational surface is tracked as its own engine, because it answers differently from the Overview.
- Sources alongside.
- Which pages the Overview drew on, so you can see whether it is your page or a competitor's doing the work.
Google · “ai search visibility tool”
AI Overview
AI search visibility tools measure how often assistants such as ChatGPT and Gemini mention a brand. Options include AI Mention Tracker, which reports each engine separately with sample sizes…
aimentiontracker.aisearchengineland.comg2.com
Collected from the live result page, so it moves when Google moves.
Everything we measure, on every plan
Seven measurements, each one tracked continuously across all six engines. No feature is held back for a larger plan — see everything we measure, and what we deliberately do not.
AI visibility tracking
How often each engine names you, as a rate with its sample size.
Sentiment analysis
How you are described when you are named, with the evidence.
Brand monitoring
Six kinds of change watched for, each emailed once.
Citation tracking
Which pages get cited, including when you are cited but not named.
Share of voice
Your share of the answers, against whoever is named beside you.
Prompt tracking
Pick the prompts to track, with search volume beside each one.
Answer archive
Every answer kept in full, with its citations, and searchable.
Common questions
If something here is unclear, ask us before you sign up rather than after.
What is an AI mention tracker platform?
It measures how often an AI assistant names your brand when someone asks a question in your category, and which web pages the assistant cited to build that answer. Traditional rank tracking measures position in a list of links. AI answers do not have positions in that sense, so visibility is measured as a share of answers rather than a rank. A platform, rather than a one-off checker, runs those questions on a schedule and keeps the history.
How does it help me get mentioned more?
By telling you which of three problems you have. Never retrieved means the assistant is not reading your pages, which is a coverage and indexing problem. Retrieved but not cited means your page lost to a better one. Cited but not named means the assistant read you and credited someone else, which is a framing problem on the page itself. Each has a different fix, and a blended score cannot tell them apart.
How is this different from tracking Google rankings?
A Google result is a stable, ranked list. An AI answer is generated fresh, varies between runs, and often cites sources it never names in the text. So a single check tells you almost nothing. We run the same question repeatedly and report a rate with the number of answers behind it, which is the only honest way to describe something that varies.
Which AI assistants do you track?
ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode. We record how each answer was collected, because a scraped product surface and a model API answer are different measurements and should never be averaged into one score.
Can you be cited by an AI assistant without being mentioned?
Yes, and it is more common than most brands expect. The assistant reads your page, uses what it learned, then recommends someone else by name. It means your content is good enough to retrieve but not framed in a way the model repeats. We surface this as its own state rather than hiding it inside a single score.
How many answers do you need before the number means anything?
We show a fraction rather than a percentage below five answers, because one run is a draw, not a rate. Percentages appear once there is enough sample to support one, and every figure on every screen carries its denominator.
Do I need a card to start?
Yes. The trial is 7 days and needs a card, and nothing is charged during those 7 days. Cancel before the end and you pay nothing. We say so on the button rather than leaving you to find out inside a payment form.
From the blog
Writing on measuring AI visibility honestly.
Are AI Overviews accurate?
Google says AI Overviews rarely hallucinate, and that holds up. It is not the accuracy question that affects your brand, and the difference is measurable.
See where you stand
Connect a brand, pick the questions your buyers ask, and get your first snapshot in a few minutes. Card required, nothing charged for 7 days.