If you're asking whether AI Overviews reduce clicks, the honest answer is yes, on the searches where they show up, and the drop is bigger than most people expect. The strongest evidence we have, a randomized field experiment, found outbound organic clicks fell by 39.8% when an AI Overview appeared on the results page. That's the number to anchor on. What the data does not support is the version of this claim that spreads on social media: that every website loses half its organic traffic because of AI Overviews. That version mixes up several different measurements, and once you sort them out, a clearer picture shows up.
Call the evidence grade this: strongly supported for a real drop in clicking behavior once an AI Overview appears, and mixed for what that means for a site's total organic traffic. Both things can be true at once, and the rest of this piece walks through why.
What "click loss" actually means
Before you can weigh the AI Overviews CTR impact on your own site, it helps to know that the studies making headlines aren't all measuring the same thing. A few terms get used interchangeably in coverage of this topic, and they shouldn't be.
Organic CTR is clicks divided by impressions for a result or a group of queries. A drop in that number doesn't automatically mean the same percentage drop in your total traffic, because it's calculated per search, not per site.
Position-one CTR is the click rate for the top traditional organic result specifically. This is the number Ahrefs has tracked, and it's especially sensitive to an AI Overview sitting above it on the page.
Outbound organic clicks are clicks that leave Google for an external site. This is the closest thing to a direct answer to "do AI Overviews reduce clicks that actually reach my site," and it's what the randomized experiment measured.
Publisher referral traffic is what a defined group of sites receives from Google Search over time. It can show a real business hit for those specific publishers without telling you anything about the web as a whole.
Aggregate organic click volume is the total clicks Google says it sends out across every site. This can hold steady even while individual publishers, categories, or query types are losing ground, because traffic can shift between sites rather than disappearing.
Citation clicks are clicks on a source link inside the AI Overview itself. A page can get a small citation-click bump compared to a page that isn't cited, while both still receive fewer total clicks than a normal results page would have sent.
Keep these six apart as you read, because the disagreement between studies is mostly a disagreement about which one they measured.
The evidence for a real drop in clicks
Start with the strongest piece of AI Overviews click loss data available: a field experiment by researchers Saharsh Agarwal and Ananya Sen, posted on SSRN in April 2026 and revised in July 2026. Users were randomly assigned through a Chrome extension to see either standard Google Search with AI Overviews or a version with them removed. Random assignment is what lets a study claim cause and effect instead of just correlation.
The results: outbound organic clicks fell by 39.8% when an AI Overview appeared, and zero-click searches rose 34.5%. Sponsored clicks and overall search frequency didn't change, and for users who did click through, engagement with the destination site afterward looked the same either way. This is the one study in the mix that can fairly be called causal, though the published summary doesn't state its sample size or exact observation window, so treat the percentage as a solid estimate rather than an exact figure to quote at decimal precision.
Pew Research Center backs this up with real observed browsing behavior, published July 22, 2025. It tracked 900 U.S. adults across 68,879 Google searches, including 12,593 that produced an AI summary. Traditional result clicks happened in 8% of visits when a summary appeared, versus 15% when it didn't. Only 1% of visits to pages with a summary included a click on a link inside it. And 26% of sessions ended right after a summary appeared, compared to 16% without one. Pew didn't randomly assign anything, so it can't prove the summary caused the difference the way the field experiment can, but it shows the same pattern in real search behavior.
Ahrefs has run the largest observational comparison, twice. Its April 2025 study pulled 300,000 keywords, split evenly between ones that triggered an AI Overview and similar informational keywords that didn't, then compared Google Search Console data from before and after the U.S. rollout. Position-one CTR for the AI Overview group fell from 7.3% to 2.6%, an estimated 34.5% lower than the informational cohort's own decline over the same period. Its February 2026 update, using December 2023 versus December 2025 data, found a steeper drop: position-one CTR for AI Overview keywords fell to 1.6%, against an estimated counterfactual of 3.7% without an AI Overview, which works out to roughly 58% lower. That update also found the effect wasn't limited to the top spot: estimated reductions ranged from about 51% at position two down to 19% at position ten.
Seer Interactive has tracked its own client accounts across several reports since September 2024, and the pattern holds there too. Its first report, covering more than 100 clients and 7,800 queries, found organic CTR at 2.94% without an AI Overview versus 0.84% with one, close to a 70% decline. Its most recent update, covering 53 brands and 5.47 million tracked queries through early 2026, found CTR for non-cited brands in AI Overview results down 67% over 2025. None of these figures are the same thing as a formal citation rate calculation for a single brand, which is a narrower, repeatable measurement you'd run on your own account rather than a study-wide average.
Two publisher-level datasets round this out. Authoritas, reporting on 3,500 UK news search terms in July 2025, found publisher CTR down 47.5% on desktop and 37.7% on mobile when an AI Overview appeared. Digital Content Next, surveying 19 publishers over eight weeks in mid-2025, found median year-over-year Google referral traffic down 10% overall, with individual publishers in its report describing CTR falling from 5.1% to 0.6% on affected pages.
The evidence against a total collapse
Here's where the picture gets more complicated, and where a lot of the alarming headlines overreach. Google said, in an August 6, 2025 company statement, that total organic click volume from Search to websites was relatively stable year over year, and that average click quality had gone up slightly, meaning fewer users bounced straight back to the search results. Google also said traffic was shifting between sites, with some gaining and others losing, rather than collapsing across the board.
That claim is real counterevidence to the idea of a universal, web-wide traffic collapse. But it comes with an important caveat: Google didn't publish a sample size, a methodology, an exact percentage, or any breakdown by query type, ranking position, or AI Overview exposure. It's a first-party statement, not a reproducible study, so it deserves attention without being treated as more rigorous than the studies it's pushing back against.
The two claims aren't actually as contradictory as they sound once you notice they're measuring different levels. Google is describing aggregate click volume across the entire web. Ahrefs is describing estimated CTR for a single ranking position. Pew is describing what a searcher does right after visiting a results page. Seer is describing CTR inside a defined set of client accounts. Authoritas and Digital Content Next are describing referral traffic for specific groups of publishers. A stable aggregate total can sit comfortably alongside real losses for individual sites, if the clicks that used to go to one publisher are now landing on a different site, a different page, or a new search entirely instead of vanishing.
What primary sources and methods actually establish
It's worth being precise about which of these findings is actually causal and which is correlational, because the studies get flattened together in most coverage of AI Overviews organic traffic studies, and that flattening is where the exaggerated claims come from.
Across the AI Overviews click loss data reviewed here, only the Agarwal and Sen field experiment can fairly claim to show cause and effect, because it randomly assigned users to see AI Overviews or not. Everything else, Ahrefs, Seer, Pew, Authoritas, and Digital Content Next, is observational. These studies compare groups of queries or sites that already differ in other ways: ranking position, query mix, seasonality, brand authority, and the general drift of other Google features like local packs and featured snippets that also eat into organic clicks. Ahrefs itself has noted that even its non-AI-Overview comparison keywords lost some CTR over the same period, which points to other forces at work alongside AI Overviews specifically.
Google's own statement sits in a third category: first-party and directionally informative, but not independently checkable. Sundar Pichai was also quoted in a May 2025 report saying links inside AI Overviews get higher CTR than the same links would outside one, which isn't actually inconsistent with the independent research. A page cited inside the summary can do better than an uncited one on the same results page, while the page overall still sends fewer people to the open web than a results page without an AI Overview would have.
One more wrinkle worth knowing before you read any single number as gospel: Authoritas separately found that roughly 70% of the pages appearing in AI Overviews changed over a two to three month window, more churn than ordinary organic rankings show. That volatility makes any snapshot study, including the ones cited here, a picture of one moment rather than a fixed and permanent state.
Why the percentages don't agree
Once you've read six studies reporting six different numbers, from 34.5% to 70%, it's tempting to assume someone is wrong. Usually nobody is. The studies disagree because they differ on:
- What's in the denominator. Impressions, search visits, queries, publisher sessions, and total site traffic are not interchangeable bases for a percentage.
- What's being measured. One ranking position, one search, one keyword group, one client account, or one publisher's overall referral number all produce different pictures of the same underlying shift.
- The comparison group. Similar keywords without an AI Overview, the same keywords before the feature rolled out, and users who never saw a summary are three different baselines.
- The time period. The gap between Ahrefs' 34.5% estimate in April 2025 and its 58% estimate in February 2026 is partly a real widening effect and partly just a different snapshot in time.
- Citation status. A cited page consistently does better than an uncited one on the same query, so lumping cited and uncited results together will blur the real size of the effect either way.
Given all that, treat every individual percentage as a description of one specific dataset rather than a universal constant you can apply to your own site. That's the real takeaway from reading the AI Overviews CTR impact literature side by side: the direction is consistent, the size is not.
The verdict and what to check in your own data
So, do AI Overviews reduce clicks? Yes, on the searches where they appear, and the causal estimate lands around 40%, with several observational studies of top-ranking pages reporting steeper declines depending on the cohort and time period. Reading across the AI Overviews organic traffic studies as a set, that's the honest summary: a real, consistent drop in clicking behavior, without a single confirmed number for what it costs every site. What isn't established is a single number you can multiply by your own traffic to get your loss. Aggregate web traffic looks roughly stable, publisher-level datasets show real losses, and both of those can be accurate at the same time because clicks are shifting around rather than disappearing uniformly.
If you want to know what's actually happening to your own pages, total organic traffic alone won't tell you. Google Search Console doesn't cleanly separate AI Overview impressions and clicks from the rest of your Search data, so a single account-wide number can hide as much as it reveals. What actually helps is pulling apart your queries where an AI Overview shows up from the ones where it doesn't. On the ones that do, check whether you're being cited inside the summary, then watch how those two groups move separately over time instead of staring at one blended CTR figure. That's the only way to tell whether a dip you're seeing is an AI Overview effect on a specific set of queries, or something else in your rankings entirely.
What would change this verdict
A few developments would shift where this evidence lands. If Google published a real methodology behind its "stable clicks" claim, with a sample and a breakdown by query type and ranking, that would either confirm or seriously complicate its counterevidence. If more studies replicate the field experiment's randomized design across different query types and countries, the causal estimate would firm up well beyond one paper. And Seer's own early-2026 numbers already show a partial rebound in CTR after 2025's steep decline, which is a reminder that these effects move as Google's interface and user habits keep changing, not a fixed rule set in stone.
None of this is a recovery guide, and that's on purpose. If your own citations disappeared after a Google model update, that's a specific diagnosis and rebuild process on its own. And refreshing a specific page to try to recover lost citations is a different, narrower fix again. This piece only covers whether the click loss is real and how big the honest range is.
