How to measure AI visibility in 2026, and track AI Overviews

The verdict

1. AI visibility measured and proven

   supereoour pick, and we make it

   • best for teams that want AEO done, not a dashboard: supereo does everything it takes to get you recommended, from your site and pages to the roundups, placements and videos AI reads;
   • it proves each move against control questions, with sample sizes and ranges on every number;
   • run it from Claude or ChatGPT: an MCP server with the same tools supereo's own agent uses.

   From $149 a month

   Run the free auditFree, no card, every answer shown.

The steps

1. Write the questions your buyers ask

   Start from what people search and ask, not from your brand name: the problems, comparisons and alternatives your buyers look up. The real run below reads monthly searches for this page's own phrases.

2. Ask every engine, many times

   Ask ChatGPT, Gemini, Perplexity, Claude, Google's AI Mode and AI Overviews the same questions on a schedule. One check is one draw from a moving answer^[1].

3. Count named, recommended and cited apart

   For every answer, record whether it names you, recommends you and cites a page on your site^[7].

4. Put a range on every rate

   Report each rate with how many answers it rests on and the range it likely sits in. A change inside the range is noise^[3].

5. Open Search Console's generative AI report

   Check your impressions and pages in AI Overviews and AI Mode in Search Console's Search Generative AI performance report^[10].

6. Record the AI Overviews themselves

   Search your key queries and keep each AI Overview and the links it cites. Google shows an overview only when it adds something^[12], so log whether one showed at all.

7. Count the visits AI sends you

   In your analytics, count visits that arrive from AI apps by their referrer or their utm_source tag, and in your server logs the AI crawlers that read your pages, checked against each company's published addresses^[15].

8. Test each change against controls

   Before you change a page, pick target questions it should move and control questions it can't, then compare how much each group moved^[8].

What gets cited

Real runs

{
  "mode": "volumes",
  "keywords": [
    "how to measure ai visibility",
    "how to track ai overviews",
    "ai visibility tracking",
    "ai overview tracker",
    "ai visibility score"
  ]
}

{
  "mode": "volumes",
  "location": "United States, English",
  "volumes": [
    {
      "keyword": "ai overview tracker",
      "monthlySearches": 390,
      "cpc": 41.9,
      "competition": 0.04,
      "difficulty": 6,
      "intent": "informational"
    },
    {
      "keyword": "ai visibility score",
      "monthlySearches": 210,
      "cpc": 34.88,
      "competition": 0.36,
      "difficulty": 31,
      "intent": "informational"
    },
    {
      "keyword": "ai visibility tracking",
      "monthlySearches": 320,
      "cpc": 34.05,
      "competition": 0.2,
      "difficulty": 25,
      "intent": "informational"
    },
    {
      "keyword": "how to measure ai visibility",
      "monthlySearches": 90,
      "cpc": 34.9,
      "competition": 0.08,
      "difficulty": 13,
      "intent": "informational"
    },
    {
      "keyword": "how to track ai overviews",
      "monthlySearches": 90,
      "cpc": null,
      "competition": 0.01,
      "difficulty": 34,
      "intent": "informational"
    }
  ],
  "notFound": [],
  "costCents": 1.27008
}

Why one check can't measure AI visibility

Ask ChatGPT the same question twice and you can get two different lists. Ahrefs found that half or more of the citations it tracked were gone the next day and that most of those came back within weeks, and its advice is not to judge AI visibility from one check^[1]. SE Ranking ran the same queries through Google's AI Mode three times and found it often picked completely different pages each time^[2].

So a screenshot of one answer, or a dashboard that asks each question once a day, measures luck as much as visibility. The fix is the one any survey uses: ask the same questions many times, count the answers, and say how sure the count is.

That also changes what a good number looks like. A rate with its sample size and its range tells you whether last week's rise was real. A single number with neither can't tell you that, however precise it looks.

How many answers you need

The fewer answers behind a rate, the bigger a difference has to be before you can trust it. Graphite worked this out for AI answers: with few answers only large gaps show, and it takes many more to separate two rates that are close^[3].

supereo publishes the numbers it runs on. A test of one change asks 10 target questions it should move and 10 control questions that it shouldn't^[4], with 15 runs of every question before the change^[5] and at least 25 of each after it^[6].

You don't need those exact numbers to measure your own visibility. You need enough answers per question that the range around each rate is narrower than the change you're looking for, and the same questions every time.

Named, recommended and cited are three different numbers

An answer can mention your brand without recommending it, and recommend you without citing your site. supereo's judge makes those three calls on every answer: named, recommended and cited^[7].

Track them apart. Being named shows the engine knows you. Being recommended is the one that sells. Being cited shows which of your pages the engine reads, which tells you where to make changes.

The metrics worth tracking

Lead with how often answers recommend you, since that's the one that sells. Keep beside it how often they name you, which shows the engine knows you, and how often they cite your pages, which shows the pages it reads^[7].

Keep the cited pages themselves, not only the rate: which of your URLs get cited, and which other pages get cited where yours could be. That list is where the work starts.

Then the measures from outside the answers: your impressions in Google's AI features from Search Console^[10], visits from AI apps in your analytics, and AI crawlers' reads in your logs. They move more slowly, and they show the result in traffic.

There's no standard good rate: it depends on your market and on who you're up against. Compare your rate with the competitors named in the same answers, and watch whether it moves beyond its range.

How often to measure

Often enough to have answers to count, and on a fixed schedule. supereo asks every question on six engines three times a week, and never twice on one engine within half an hour^[16].

Read rates over weeks rather than days. supereo's rates cover the last four weeks of answers^[17], so one odd day can't swing them, and each comes with its range.

How to track AI Overviews and AI Mode

Google now gives you this data directly. Its Search Generative AI performance reports in Search Console show your impressions inside AI Overviews and AI Mode, on their own, and since August 2026 every website has them^[9]. They count how often your pages appeared in those features and which pages did^[10].

The same impressions also count in Search Console's overall performance report, under the Web search type^[11], so a rise there can come from classic results or from AI features. The dedicated report tells them apart.

What Search Console doesn't show is the answer itself: which other sites the AI Overview cited, and what it said about you. For that, search your buyers' queries and record each overview and the links it cites, again and again, since Google shows an overview only when it adds something and often doesn't^[12].

How to increase AI visibility once you've measured it

A measurement is worth its next step. Read the gaps engine by engine: where an engine never names you, where it names you but recommends someone else, and where it cites other pages for questions your own pages answer.

Then close them. Let AI crawlers read your site, answer each buyer question first on your own pages, and get your brand onto the pages the engines cite. Each engine has its own way in: how to get recommended by Claude, how to show up in AI Overviews and AI Mode and how to get cited by Perplexity. Then ask the same questions again, so you can see what each change did.

If you'd rather a tool kept count, pick one that shows sample sizes and ranges, asks every engine your buyers use, and tells named, recommended and cited apart. Our guide to the best AI visibility tools compares them, with supereo as our pick.

Measuring a change, not just a level

Knowing your rate is the start. Knowing whether a change you made moved it is the hard part, because everything else moves too: a model update, a competitor's launch, a slow week.

Control questions handle that. Pick questions your change can't affect, ask them alongside the ones it should, and compare: how much the targets moved minus how much the controls moved^[8]. That is how supereo grades every change it makes, and why its results come with a range.

Do the same by hand if you like, or let supereo do it for you: its free audit shows where you stand today on five AI engines, with no card. Run it here, and see How supereo measures for every rule its tests follow.

Have it done: supereo

Questions

How we compiled this

Where this page's words say how an AI engine finds, reads or cites pages, the sentence links the engine's own documentation, or another source, and the day it was read.

What each engine cites comes from the supereo Index's newest run: the same buyer questions asked once on each chat engine, and their search keywords on Google AI Overviews, in each of the Index's topics, with every page each answer cited kept. For each engine and topic the run keeps the ten pages its answers cited in the largest number of answers, and how many cited each, and this page counts those pages by type.

A page's type comes from its address alone: a Reddit, YouTube, Wikipedia or GitHub page by its site, a home page when its address has no path, a pricing page when its address ends in /pricing, a docs or help page when it is on a docs, developers, help, support or learn site, or under /docs, /help or /hc, a list, comparison or review when its address names best, top, alternatives, vs, versus, compare, review or tools and every other page an article or other page. An AEO tool's own page counts in its type but is never named as an example, and neither is a search engine's results page; in AI visibility tools, the topic supereo competes in, the only examples are the AI engines' makers' own docs and help pages. The findings are about what gets cited, not which tool.

The runs are real calls to supereo's MCP server on its staging environment, made as a house account, recorded as they happened: the arguments as sent, each answer as the client received it, its JSON indented and trimmed only where it says so, the time each call took and when it ran.

We make supereo, so it is the one tool this page names as a pick.

How supereo measures its own results, with every number it runs on: How supereo measures.

supereo (supereo.app) is an AI agent that does everything it takes to get your brand recommended by ChatGPT, Gemini, Perplexity, Claude and Google, on your site and off it: site fixes, new pages, roundups, placements and video. It proves what worked against control questions. Made by superagnt.

Sources

1. Don't judge AI visibility from one check. 50–83% of citations are gone the next day, but about 75% of those come back within 20 days. Count how many of the last seven days you were cited. ahrefs.com, read 7 November 2026.
2. The average overlap of exact URLs between the three datasets was just 9.2%. This means that AI Mode often chooses completely different pages each time, even when the query doesn’t change. seranking.com, read 7 November 2026.
3. For example, with 10 responses, we can statistically distinguish between 75% and 25%, but not 75% and 50%. With 50 responses, we can statistically distinguish between 50% and 30%, but not 50% and 40%. graphite.io, read 7 November 2026.
4. 10 target questions, the ones the change should move, and 10 control questions from your panel it shouldn't supereo.app, read 7 November 2026.
5. A baseline: 15 runs of every question before the change supereo.app, read 7 November 2026.
6. After the change: a run of every question about every 42 hours, at least 25 of each before a verdict. supereo.app, read 7 November 2026.
7. Named: the answer mentions you. Recommended: the answer offers you as something to buy or use, with the answer's own words kept as its quote. Cited: the answer cites a page on your site. supereo.app, read 7 November 2026.
8. Difference in differences: how much the targets moved, minus how much the controls moved. supereo.app, read 7 November 2026.
9. Note: As of August 31, 2026, we've rolled out these insights to all websites worldwide. Today, we're excited to announce the launch of new Search Generative AI performance reports in Search Console, including dedicated reports for Search and Discover, to help you understand your site's visibility within generative AI features on Search. developers.google.com, read 7 November 2026.
10. Impressions: How often URLs from your site appeared in generative AI features in Search and Discover. Pages: Check which URLs appeared within AI features. developers.google.com, read 7 November 2026.
11. Just like the rest of the search results page, sites appearing in AI features (such as AI Overviews and AI Mode) are included in the overall search traffic in Search Console. In particular, they're reported on in the Performance report, within the "Web" search type. developers.google.com, read 7 November 2026.
12. AI Overviews are only shown when our systems determine that it is additive to classic Search, and as such, often don't trigger. developers.google.com, read 7 November 2026.
13. Both AI Overviews and AI Mode may use a "query fan-out" technique developers.google.com, read 7 November 2026.
14. To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet, fulfilling the Search technical requirements. There are no additional technical requirements. developers.google.com, read 7 November 2026.
15. If a crawler has a source IP address on this list, it indicates that the crawler is coming from Anthropic. support.claude.com, read 7 November 2026.
16. Every question runs on six engines, 3 times a week each, and never twice on one engine within 30 minutes. supereo.app, read 7 November 2026.
17. A rate is the share of the last 4 weeks of answers that recommend you, shown as answers out of every 100. supereo.app, read 7 November 2026.
18. With expanded AI Overviews, more planning and research capabilities, and AI-organized search results, our custom Gemini model can take the legwork out of searching. blog.google, read 7 November 2026.
19. AI Overviews can take the work out of searching by providing an AI-generated snapshot with key information and links to dig deeper. support.google.com, read 7 November 2026.
20. Features cannot be turned off. However, you can select the Web filter after you perform a search. This filter displays only text-based links without features like AI Overviews. support.google.com, read 7 November 2026.
21. To develop and improve generative AI experiences in Search and the related machine learning technologies, Google uses people's interactions with Search and these AI experiences. support.google.com, read 7 November 2026.
22. visibility in AI search results across ChatGPT, Gemini, AI Mode, and AI Overviews. semrush.com, read 7 November 2026.
23. Yes, the AI Visibility Checker is free to use. You can use the tool three times per day for free, without registering or creating an account. semrush.com, read 7 November 2026.
24. Free AI Visibility Checker See how your brand shows up in AI search results from ChatGPT, Gemini, Perplexity, Copilot, and Google AI Overviews. No signup required. Results in seconds. ahrefs.com, read 7 November 2026.