For twenty years, rank tracking rested on a simple idea: type a query, count down the list, write down a number. Google’s AI Mode does not return a list to count down. It returns a written answer with a handful of links beside it, and the next person who asks may get a different one. For teams that monitor search visibility, that changes what a ranking means and how it should be measured.
This guide covers what AI Mode is, what Google’s own reports count, what independent studies say about how stable its citations are, and a measurement approach that holds up.
What AI Mode is
AI Mode is the conversational search tab Google opened in Labs in March 2025 and rolled out to everyone in the US in May 2025. By October 2025 it was available in over 200 countries and territories, and at Google I/O in May 2026 the company said it had passed one billion monthly users. Its queries are, in Google’s words from the October 2026 earnings call, “on average three times longer than traditional searches”.
It is not the default. In June 2026, Google’s VP of Search Engineering wrote: “We’re not planning to make AI Mode the default for Chrome searches”. But the boundaries are blurring. A question asked from an AI Overview can now continue in AI Mode, and in October 2026 Google began scaling ads in AI Mode beyond the US, with “one seamless experience across AI Overviews, AI Mode, and traditional Search” as the stated goal.
Why position stops working
Three things about how AI Mode builds an answer break the old model.
It runs many searches, not one. Google describes a “query fan-out” technique: the system splits a question into subtopics and searches for each one at the same time. Your page is competing in searches you never typed, so the keyword you track may not be the one that earned the citation.
The answer is written fresh. Links appear beside and inside a generated response, not in fixed slots. Google also notes that when it lacks confidence in an AI response, AI Mode shows a set of web links instead.
The conversation moves on. Follow-up questions are, in Google’s words, “essentially performing a new query”. A single session can produce several answers, each with its own sources. Opt-in personal context, drawn from search history and connected apps, can shape answers for signed-in users too.
What Search Console counts
Search Console does record AI Mode, but it blends it into figures you already know.
| What | How Google counts it |
|---|---|
| AI Mode clicks and impressions | Counted in the Performance report totals since June 2025, under the Web search type |
| Position in AI Mode | ”follows the same methodology as a Google Search results page” |
| Clicks in AI Mode | A click on a link to an external page counts as a click |
| Follow-up questions | Treated as a new query |
| AI Overviews | The overview takes one position, and all its links share it |
| Generative AI reports (June 2026) | Impressions only, by page, country, device and date |
The generative AI reports are the closest thing to a direct view. They show how often your pages appeared in AI Overviews and AI Mode, but not clicks, positions or the queries that triggered them. Google says they reached all websites worldwide by 31 August 2026. There is still no filter to separate AI Mode clicks from classic results in the main report.
So Search Console tells you whether you are showing up in AI features, and where that traffic lands. It does not tell you which questions you are cited for, how often, or who is cited instead.
What the studies show
Independent data on AI Mode comes mostly from SEO tool companies, which sell visibility tracking, so read the numbers as directional. They point the same way.
| Finding | Source |
|---|---|
| 14% of AI Mode’s cited URLs, on average, were also in the organic top 10, and 21.9% of domains | SE Ranking, 10,000 US keywords, June 2025 |
| The same keywords, run three times in one day, shared only 9.2% of exact URLs | SE Ranking, same study |
| 58% of sites cited by AI Mode on day one were gone by day two, and 6.8% lasted all 20 days | Ahrefs, 1,996 questions, October 2026 |
| AI Mode and AI Overviews cited the same URLs for the same query 13.7% of the time | Ahrefs, 540,000 query pairs, September 2025 data |
Clicks are scarcer too. Pew Research, which is not a tool vendor, found that people clicked a traditional result on 8% of visits when an AI summary appeared, against 15% without one, and clicked a link inside the summary on 1% of visits. Google’s position is that total organic clicks have been “relatively stable year-over-year”, with “slightly more quality clicks”, though it has not published figures.
The practical reading: ranking well still helps, but it does not predict a citation, and one snapshot of an answer is close to noise. Ahrefs put it plainly: “Don’t judge AI visibility from one check.”
What to track instead
Treat AI Mode as a probability, not a position. These measures hold up better than a single rank.
- Citation rate. Out of N runs of a question, how often is your domain cited? Report it with the number of runs behind it.
- Share of citations. Across a set of questions, what share of all citations go to you and to each competitor?
- Persistence. How many days in a row does a citation survive? A page that is cited on 15 of 20 days is worth more than a page cited on one lucky day.
- Which page is cited. A citation to your blog and a citation to your pricing page are not the same win. Track the URL, not just the domain.
- The classic baseline. Keep tracking organic positions. They are still the best leading signal, and our guide to accurate keyword rankings covers doing that cleanly.
- Clicks and conversions. Search Console and your analytics show what the visibility is worth, even if they cannot show where it came from.
Track questions, not just keywords. AI Mode queries are long and phrased as tasks, so build the question set from real customer language: support tickets, sales calls and the questions on your own FAQ.
How to collect it
Because answers vary, collection design matters more than ever.
- Repeat each question. Run each question several times, on several days, and store every run. A rate needs a denominator.
- Pin the variables. Fix language, device and location for each series, and use a clean, logged-out session so personal context does not leak in.
- Measure location. Google does not say how much AI Mode answers vary by place, and Search Console reports AI impressions by country, so test it. A geo-grid works for local questions, and residential proxies let you see the answer a local visitor would see in each market.
- Store the raw answer. Keep the full text and every link, with a timestamp, so you can re-score runs when your method changes.
- Separate noise from change. A citation that vanishes for a day is normal. A drop in citation rate across many questions and days is a signal. Our guide to detecting SERP volatility covers the baselines.
Keep it to public, signed-out results, respect Google’s terms and rate limits, and query no harder than you need. Our earlier piece on monitoring AI search covers the responsible-use side in more detail.
What site owners can control
Google says there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”. The usual preview controls (nosnippet, data-nosnippet, max-snippet and noindex) apply. In June 2026 Google also began testing a Search Console setting to opt a site out of its generative AI features entirely, with the warning that sites that opt out “will not receive traffic or impressions” from them. Weigh that trade before using it.
The bottom line
AI Mode turns a ranking into a rate. Search Console will tell you that you appeared and what the clicks were worth, but not which questions you win or how reliably. Build a question set, run it repeatedly from fixed and local vantage points, store every answer, and report citation rates with their sample sizes. Keep classic rank tracking running beside it: it is still the best early signal for whether you will be cited at all.
Sources and references
- Google, Expanding AI Overviews and introducing AI Mode (5 March 2025): blog.google
- Google, AI Mode in Google Search: updates from Google I/O 2025 (20 May 2025): blog.google
- Google, AI Mode expands to more languages and locations (7 October 2025): blog.google
- Google, Search at I/O 2026 (19 May 2026): blog.google
- Google, Alphabet Q3 2026 earnings remarks (28 October 2026): blog.google
- Google, Ads in AI Mode expand globally (28 October 2026): blog.google
- Search Engine Roundtable, Google not planning to make AI Mode the Chrome default (June 2026): seroundtable.com
- Google Search Help, AI Mode: support.google.com
- Search Console Help, Performance report metrics: support.google.com
- Google Search Central, Search generative AI performance reports (3 June 2026): developers.google.com
- Google Search Central, AI features and your website: developers.google.com
- Google, New controls for website owners (3 June 2026): blog.google
- Google, AI in Search is driving more queries and higher quality clicks (6 August 2025): blog.google
- Pew Research Center, Google users are less likely to click on links when an AI summary appears (22 July 2025): pewresearch.org
- SE Ranking, AI Mode research (29 August 2025): seranking.com
- Ahrefs, How stable are AI citations? (29 October 2026): ahrefs.com
- Ahrefs, AI Overviews vs AI Mode (15 December 2025): ahrefs.com