- How often an AI Overview appears depends on who counts: 48% of BrightEdge’s tracked queries in February 2026 [8], 43% of US searches in Similarweb’s 2026 report [9], and above 20% of real searches in SparkToro’s clickstream panel. [10] Keyword trackers and clickstream panels count different populations.
- With an AI summary present, people click a result 8% of the time, versus 15% without one. Clicks on links inside the summary: 1%. [1]
- Zero-click reached 68.01% of US Google searches in January to April 2026, against 60.45% in 2024. [10] They also cut clicks to the #1 organic result by 58%. [2]
- The clicks that survive are commercial and bottom-of-funnel, the ones that were going to convert. [5]
- Stop reporting sessions. Report revenue per visit, or the numbers will lie to you.
Google stopped sending the click and kept the query. That is the whole story, and every argument about SEO in 2026 is a footnote to it. The search still happens (someone still wants the answer), but the answer now lives on Google’s page, in a box you did not write and cannot rank inside.
The scale is no longer arguable.
Pew Research put numbers on the behavior rather than the ranking. Across 900 US adults, in searches that surfaced an AI summary people clicked through to a traditional result only 8% of the time. In searches without one, 15%. And the summary’s own citations, the links Google offers as the consolation prize, earned a click in roughly 1% of visits. [1][3]
15% click a result. 16% end the session there. [1] The page still does its job.
8% click a result. 26% end the session there. [1] The answer already happened.
Read those two columns as a behavior change, not a ranking change. The page you rank on didn’t get worse. The reason to visit it got answered before the visit. That distinction decides what you are actually fighting. You are not losing to a competitor who out-optimized your title tag. You are losing to a summary that made the trip unnecessary, and no amount of on-page work reopens a tab the reader closed because they already got what they came for. The lever moved from the page to the query: some queries still force a click, and most no longer do.
The decline is measured, repeatedly, by people who disagree about everything else
This is not one vendor’s talking point, and the way to check that is to open the studies rather than a roundup of them. Start with the number people quote most and get wrong most, because how often an AI Overview appears depends entirely on who is counting and what they are counting. On Advanced Web Ranking’s 8,000-keyword US tracker, read on March 2, 2026, it was 65.07% of personalized results and 49.43% of non-personalized ones: the same keywords, the same day, sixteen points apart. [4] BrightEdge put it at about 48% of its tracked queries in February 2026, up from about 30% a year earlier. [8] Similarweb’s 2026 report says 43% of US searches, and does not publish the method behind it. [9] SparkToro, reading Similarweb’s clickstream panel of searches real people actually ran between January and April 2026, says more than 20%. [10]
Those are not four answers to one question. A keyword tracker asks what share of a chosen term list triggers the feature; a clickstream panel asks what share of the searches people actually ran had one. The lists are weighted toward the informational queries that trigger AI Overviews and real search behavior is not, which is most of why the tracker numbers run two to three times the panel numbers. Take the number with its measurer, its date and its method attached, or it means nothing. What the datasets do agree on is the direction, and on what happens to the click.
| Study | What it measured | Click impact |
|---|---|---|
| Pew Research, March 2025 | Result clicks, with vs without an AI summary, 900 US adults | 8% vs 15% [1] |
| Ahrefs, December 2025 | Clicks to position #1 with an AIO present, 300k keywords | -58% [2] |
| Seer Interactive, June 2024 to September 2025 | Organic CTR on AIO queries | 1.76% to 0.61%, -61% [7] |
| SparkToro, January to April 2026 | Share of US Google searches ending with no click | 68.01%, against 60.45% in 2024 [10] |
The studies disagree on the exact figure and agree entirely on the direction. What none of these averages show is which clicks disappeared, and that is where the picture stops being uniformly bleak.
But it does not fall evenly, and that is the opening
The average hides the only thing worth knowing: which clicks disappeared. The ones Google absorbed are the cheap ones: “what is”, “how does”, the definitional top-of-funnel queries a summary answers perfectly. The ones it largely left alone are the expensive ones: pricing, comparisons, “best X for Y”, “demo”, “near me”. Commercial, bottom-of-funnel intent still mostly sits outside AI Overview coverage, and still requires a click, because the buyer needs to leave Google to actually buy. [5]
Your traffic got smaller and more valuable at the same time.
If you report organic sessions, AI Overviews will read as a slow catastrophe. If you report revenue per visit, the same period reads as a purge of traffic that was never going to buy. Both are true. Only one of them is the number that pays you, so make it the number on the dashboard before someone panics at the other one.
What survives is not smaller in value, only in volume. The reader who clicks past a fully formed answer to reach your page is self-selecting for intent that the answer could not satisfy: a spec to check, a price to compare, a vendor to trust. That is a better visitor than the average of 2023, not a worse one.
Google has no reason to hand the click back
It helps to be honest about the incentive, because it tells you which way to plan. Every click that leaves Google is attention Google stops monetizing. AI Overviews keep the reader on the results page longer, resolve more of the query inside Google’s own frame, and wrap that answer in Google’s own ad inventory. The zero-click search is not a side effect of the rollout. It is the design goal. So the sensible planning assumption is not “this stabilizes.” It is that coverage keeps climbing: the same 8,000-keyword tracker read roughly 25% in August 2024, 60.32% in November 2025 and 65.07% in March 2026. [4] And the click keeps compressing, toward the floor Pew measured in March 2025 and below it.
That reframes the whole job. You are no longer optimizing to win a click that the platform is actively working to prevent. You are optimizing for the two things a summary structurally cannot absorb: the query where the reader has to transact somewhere other than Google, and the brand the reader already trusts enough to type by name. Nearly everything in the surviving-traffic column is a version of one of those two. Informational content isn’t dead, but it has been demoted from a traffic asset to a citation asset: its job now is to be the source the summary quotes, which keeps you present in the answer even when it doesn’t send the visit.
What to actually do about it
- Move the key performance indicator (KPI). Make revenue-or-conversions-per-visit the headline organic metric, and demote sessions to a diagnostic.
- Reallocate to bottom-of-funnel. Pricing pages, comparisons, integration and “alternative to” content are where the click still lives.
- Concede the definitional queries. Do not spend a content budget out-ranking a summary on “what is retargeting.” You will lose, and the click was low-value anyway.
- Get cited in the summary. Being the source Google quotes preserves brand presence even when it does not send the click, and it still moves clicks: in Seer’s Q3 2025 data, brands cited in the AI Overview earned a 35% higher organic CTR than uncited ones. [7]
- Re-baseline the board. Show the traffic decline and the value-per-visit rise on the same slide, before the quarterly review does it for you.
Google did not kill SEO. It killed the era where a ranking and a click were the same thing. The publishers who struggle are the ones still counting the click. The ones who adapt are already counting what the click was worth.
Sources
- Pew Research Center · Google users are less likely to click on links when an AI summary appears in the resultsn = 900 US adults; searches conducted March 2025
- Ahrefs · Update: AI Overviews reduce clicks by 58%
- Search Engine Land · Google’s AI Overviews are hurting clicks: Pew study
- Xponent21 · Google AI Overviews hit 65%reading Advanced Web Ranking’s 8,000-keyword US AI Overview tracker on March 2, 2026: 65.07% of personalized results, 49.43% of non-personalized, against 60.32% in November 2025 and about 25% in August 2024
- Search Engine Land · Why bottom-of-funnel content is winning in AI search
- Ahrefs · AI Overviews reduce clicks by 34.5%the original April 2025 study of 300k keywords that the December update revised to 58%
- Seer Interactive · AIO impact on Google CTR: September 2025 updateorganic CTR on AIO queries down 61%, from 1.76% to 0.61%; cited brands earn 35% higher organic CTR
- BrightEdge · AI Overviews at the one-year mark: presence, size, and what they’re citingtracked keyword set, twelve months to February 2026: AI Overviews on about 48% of tracked queries, up from 30% a year earlier
- Similarweb · AI search stats 2026: market share, referral, and citation dataa 43%+ AI Overview appearance rate in the US in 2026; the public pages do not state the query sample or the method
- SparkToro · In 2026, less than one third of Google searches still send a clickSimilarweb US desktop and mobile clickstream panel, January to April 2026: 68.01% of searches end without a click against 60.45% in 2024, and AI Overviews on 20%+ of searches
Frequently asked questions
How much have AI Overviews actually reduced clicks?
By Ahrefs’ December 2025 measurement, the top organic result gets 58% fewer clicks when an AI Overview appears above it, up from a 34.5% reduction in their April 2025 study. Pew Research found people clicked a traditional result in just 8% of searches that showed an AI summary, versus 15% of searches without one.
Do people click the links inside the AI summary instead?
Rarely. Pew Research found users clicked a source link inside the AI summary in about 1% of all visits. The summary is designed to end the search, not to route it onward, so treating AI-Overview citations as a traffic channel is optimistic at best.
Are all types of search affected equally?
No, and this is the important part. Informational, top-of-funnel queries (“what is”, “how to”) are the hardest hit. Commercial and bottom-of-funnel queries (pricing, comparisons, “near me”, “demo”) still largely require a click and sit outside heavy AI Overview coverage, which is why the surviving traffic skews toward intent.
What metric should replace sessions?
Revenue per visit, or conversions per visit. Total organic sessions will fall structurally as AI Overviews expand, so a session-count KPI will read as failure even when the commercially valuable traffic is intact. Measure the value of a visit, not the volume of visits.
Is this the same thing as ChatGPT taking search traffic?
Related but distinct. AI Overviews are Google answering inside its own results page. Assistant traffic from ChatGPT and others is a separate, smaller channel. Both compress informational clicks, but AI Overviews are the larger near-term force because they affect the queries you already rank for.
Will Google eventually give the clicks back?
There is no incentive for it to. Every click that leaves Google is attention Google stops monetizing, and AI Overviews keep the reader on the results page, resolve more of the query inside Google’s own frame, and wrap the answer in Google’s own ad inventory. The zero-click search is the design goal, not a side effect. Coverage has climbed on every tracker that publishes a series: Advanced Web Ranking’s 8,000-keyword US set went from roughly 25% in August 2024 to 60.32% in November 2025 and 65.07% in March 2026, and BrightEdge’s tracked set went from about 30% to 48% in the year to February 2026. Quote any of those with its measurer and its date, because trackers and clickstream panels count different populations and the gap between them is wide.
If informational pages lose the click, are they still worth writing?
Yes, but their job changed. Informational content has been demoted from a traffic asset to a citation asset: its role now is to be the source the AI summary quotes, which keeps your brand present in the answer even when it doesn’t send the visit. Treat that citation as the new impression, not the new visit, and stop spending budget trying to out-rank a summary on definitional queries you will lose anyway.
Is the traffic that survives lower quality than before?
No, it is higher quality. The reader who clicks past a fully formed answer to reach your page is self-selecting for intent the summary could not satisfy: a spec to check, a price to compare, a vendor to trust. That is a better visitor than the average of 2023, not a worse one. Your traffic got smaller and more valuable at the same time, which is why revenue per visit tells the true story and session count does not.
Mikołaj Salecki
Writes about media, tech, and AI business for people who actually run digital. Former agency lead. Skeptic of frameworks that read better than they perform.
