AI Search

AI Overviews Cut the Top Result's Clicks by 58%, and the Hit Keeps Growing

Ahrefs looked at 300,000 keywords and found that when an AI Overview showed up, the top-ranking page got 58% fewer clicks. A year earlier, the same study measured a 34.5% drop, so the loss has grown sharply in twelve months. A separate randomized field experiment, in which real people searched with AI Overviews either shown or hidden, found that people clicked on 38% fewer websites when the AI answers were showing.

The headline number: 58% fewer clicks for position one

The headline finding is blunt. In the Ahrefs study of 300,000 keywords, the page ranking first lost 58% of its clicks when an AI Overview appeared on the results page. That is not a small dip at the margin, and it is not a rounding error in a minor metric. It is more than half of the clicks that the top result would otherwise have received, removed by a single feature sitting above it. For a site that depends on ranking first for its most valuable queries, that is the difference between a page that drives the business and a page that barely registers.

The number matters because position one has always been the prize in search. Ranking first meant the most visibility, the most clicks, and the most traffic, and whole content strategies were built around winning that spot. A 58% cut changes the arithmetic of that strategy without changing the ranking itself. The page still ranks first. It still shows up at the top of the organic results. It simply gets far fewer of the clicks that used to follow from that position, because the AI Overview answers the question before the searcher ever reaches the list of links.

How the 58% figure was measured

The Ahrefs figure comes from a keyword study, and the details of the measurement are part of why it carries weight. The study covered 300,000 keywords, which is a large enough sample to smooth out the quirks of any single query type or any single day. The comparison is between queries where an AI Overview appeared and the baseline of what the top result would be expected to receive without one. That baseline matters, because a raw click count on its own says little; the loss only becomes visible when it is measured against what position one normally earns.

The study also gives a concrete click-through rate for December 2025. For queries with an AI Overview, the position-one click-through rate was about 1.57%, around 58% below the estimated baseline of 3.73%. Those two numbers tell the same story in a different form. A baseline of 3.73% means that, without an AI Overview, roughly 37 or 38 searchers in a thousand would click the top result. With an AI Overview, that falls to roughly 16 in a thousand. The gap between those two rates is the traffic that the AI answer absorbs, and it is the traffic that publishers and SEOs are now trying to win back by other means.

A year earlier, the drop was 34.5%

The most alarming part of the Ahrefs finding is not the 58% figure on its own, but the direction it is moving. A year earlier, the same study measured a 34.5% drop for the top-ranking page when an AI Overview appeared. That was already a serious loss, and it was already enough to change how teams thought about ranking. In the year since, the loss has grown from about a third to well over half. The feature did not stay still, and neither did its effect on clicks.

That growth invites an obvious question: why would the hit get bigger over time? The study itself does not answer that question with a single cause, and it would be wrong to invent one. But the pattern is consistent with a feature that appears more often, covers more query types, and answers more of the question directly on the results page. Each of those changes, on its own, would push clicks down. Together, they explain why a 34.5% loss can become a 58% loss in a year, and why teams that planned around the earlier number now find their traffic models out of date.

The field experiment: 38% fewer websites clicked

The Ahrefs study is not the only evidence, and the second piece of evidence is in some ways stronger, because it comes from an experiment rather than an observation. In a separate randomized field experiment, real people searched with AI Overviews either shown or hidden. The randomization is the key: because the two groups were assigned by chance, the difference between them can be read as the effect of the AI answers themselves, rather than as a difference in the people or the queries. With the AI answers showing, people clicked on 38% fewer websites.

The 38% figure and the 58% figure measure different things, and they should not be confused. The 58% figure is about the top-ranking page specifically, and how many of its clicks disappear when an AI Overview appears. The 38% figure is about all websites together, and how many fewer of them get clicked when AI answers are shown. The two numbers point in the same direction, which is what makes them persuasive together: the top result loses the most, and the web as a whole loses a large share as well. The AI Overview does not only reshuffle clicks from one site to another; it reduces the total number of clicks that leave the results page.

What Pew found in real browsing: 8% versus 15%

Pew Research Center examined browsing activity from 900 US adults in March 2025, covering 68,879 Google searches, and its findings line up with the Ahrefs and field experiment results. Users clicked a result in 8% of visits with an AI summary, versus 15% without one. In other words, the chance that a search visit produced a click at all was nearly cut in half when an AI summary was present. That is a different lens from the Ahrefs study, which looked at the top result, but it tells the same story about behaviour: when the answer is on the page, fewer people feel the need to leave it.

Pew also found that only 1% clicked a link inside the summary itself. That detail matters, because it limits one of the main hopes publishers have had about AI Overviews: that being cited inside the summary would replace the clicks lost from the organic results. If only one searcher in a hundred clicks a link inside the summary, then citation is not a substitute for ranking, at least not in click terms. A citation may build awareness, and it may send some traffic, but on the evidence Pew collected, it does not send anything like the traffic that the top organic result used to receive.

Where AI Overviews show up, and where they do not

Not every query is affected equally, and the Seer Interactive findings make that clear. Seer found AI Overviews on about 36% of informational searches, but only about 5% of buying searches. The gap between those two numbers is large, and it shapes where the click loss is felt most. A site that mainly answers questions, explains concepts, or publishes guides will see AI Overviews often, and will feel the 58% effect on many of its pages. A site that mainly sells, with product pages and buying guides, will see them far less often, at least for now.

That split also explains why the industry conversation about AI Overviews can sound contradictory. Publishers of informational content report steep traffic drops, while ecommerce and buying-focused sites report far less disruption, and both can be right at the same time. The feature is concentrated where the question is informational, because that is where an AI summary can answer the query directly. Buying searches, by contrast, still tend to push the searcher toward a choice between products, and that choice still leads to a click. The risk for informational publishers is real, and the relative safety for buying pages is real too, and both follow from the same 36% versus 5% split.

The other side: AI visitors convert at about three times the rate

There is a counterweight to the click loss, and it comes from Orbit Media, which studied 97 B2B websites and found that AI visitors were about three times more likely to become leads than other organic visitors. That finding does not cancel out the 58% click loss, and it should not be read as if it does. Fewer visitors arriving is still fewer visitors, and a higher conversion rate on a much smaller base can still mean fewer leads in total. But it does change the meaning of the traffic that remains, because the visitors who do click through from an AI answer appear to arrive better informed and more ready to act.

The most useful way to read the Orbit finding is as a quality story rather than a volume story. An AI answer can filter the audience before the click happens: the searcher who reads the summary, still has a question, and then clicks is not the same as the searcher who clicks the first link out of habit. The first searcher has already done some of the research, and may be closer to a decision. For B2B sites in particular, where a single lead can be worth far more than a single visit, that filtering effect matters. Traffic is down, but the traffic that survives may be worth more per visitor than the traffic that was lost.

The traffic shift in the Allua panel

Allua recalculated Ahrefs' fixed panel of 118,671 sites, updated November 11, 2026, and the panel gives a longer view of the same shift. Between October 2025 and October 2026, Google-sourced traffic fell about 30.7%, from roughly 430.0 million to 297.8 million estimated visits. Over the same period, ChatGPT referrals grew about 43%, from roughly 3.01 million to 4.30 million. The two trends move in opposite directions, and they cover the same twelve months, which makes the comparison hard to ignore: search traffic from Google fell sharply, while referrals from ChatGPT grew sharply.

The panel is valuable because it is fixed, which means the same 118,671 sites are measured at both ends of the period. That design rules out one common distortion, where a changing sample makes it look as if traffic moved when only the sites being measured changed. Here the sites stayed the same, and the traffic still moved. The Google-sourced figure fell by roughly a third, and the ChatGPT figure rose by more than two-fifths. Both movements are large, and both happened in a single year, which is why the panel has become a reference point for anyone trying to describe the scale of the shift.

Why the ChatGPT growth covers only about 1% of the Google decline

The Allua numbers also carry a warning against a comforting misreading. The roughly 1.29 million additional ChatGPT visits were equivalent to around 1% of the observed decline in Google-sourced traffic. That is, the growth in ChatGPT referrals, large in percentage terms, is tiny in absolute terms next to the Google decline. ChatGPT referrals grew from roughly 3.01 million to 4.30 million, a gain of about 1.29 million visits. Google-sourced traffic fell from roughly 430.0 million to 297.8 million, a loss of roughly 132 million visits. The gain covers about one percent of the loss.

That arithmetic matters because it stops a tempting story in its tracks. It would be easy to say that traffic is simply moving from Google to ChatGPT, and that publishers will get it back through AI referrals. On the panel numbers, that is not what is happening. The traffic is mostly not moving; it is disappearing, because the AI answer satisfies the searcher without a click to any site. In the same October 2026 panel, search engines sent about 60 times as much traffic as identifiable AI assistants, which puts the scale in perspective: AI referrals are growing fast from a small base, while search remains the dominant source by a wide margin, even after a steep fall.

What this means for ranking first

Ranking first still matters, but it no longer means what it used to mean, and the studies in this article explain why. The Ahrefs figure shows that the top result loses 58% of its clicks when an AI Overview appears. The Pew findings show that a search visit with an AI summary is about half as likely to produce any click at all. The Seer findings show that informational queries, the kind that many content sites depend on, are the ones most likely to carry an AI Overview. Put together, those three findings describe a world in which position one is still the best organic position, but the reward for holding it has shrunk, and it has shrunk fastest for exactly the pages that used to benefit most from it.

That does not mean ranking work is pointless, and it would be a mistake to read the numbers that way. A page that ranks first still gets the clicks that remain, still gets cited in the AI Overview more often than a page that ranks tenth, and still carries the brand visibility that comes with being at the top. What has changed is the return on the investment. A team that spends the same effort to win position one now gets about half the clicks it would have received a year earlier, and that changes which pages are worth the effort, which queries are worth targeting, and how success should be measured. The ranking is the same; the payoff is not.

There is also a reporting problem hidden inside the ranking problem. Most dashboards still report position as if it were the outcome, and a page that holds position one looks healthy even while its clicks fall by more than half. That mismatch is why teams need to look at clicks and citations together, rather than celebrating a ranking that no longer delivers the traffic it once did. Position one is still worth winning, but it is no longer worth reporting on its own, because the number that used to follow from it no longer follows.

What to do instead of chasing position one alone

The practical response starts with accepting that clicks are no longer the only outcome worth tracking. The Ahrefs, Pew, and Orbit findings together point to three separate outcomes: being cited in the AI answer, being clicked, and converting the visitors who do arrive. A team that tracks only clicks will see a decline and conclude that its content has failed, even when the same content is being cited constantly and sending a smaller number of better-qualified visitors. Tracking citations, clicks, and leads separately gives a truer picture, and it stops a team from abandoning content that is actually working in a different way.

The second part of the response is about where to invest. Buying-focused queries, where AI Overviews appear on only about 5% of searches, still behave much more like traditional search, and content that serves a buying decision still earns clicks at something closer to the old rate. Informational content, by contrast, needs a different justification: it should be built to be cited, to build the brand, and to send the smaller number of visitors who still click, rather than to harvest clicks at the old volume. That is a harder brief, but it is an honest one, and it is the brief that the 58% figure, the 38% experiment, and the panel numbers all point toward.

Sources

Sources: Ahrefs (February 2026); Pew Research Center (July 22, 2025); Seer Interactive; Orbit Media; Allua (November 12, 2026).

Related reading on this site: AI Overviews Are Taking Over the Map Pack, Semrush Finds, Pharma Brands Win 9.2% of Top Rankings but Only 4.2% of AI Overview Citations, 32 AI Search Studies Reviewed: Citations Without Brand Mentions Are the Norm, Google AI Overviews Now Show Citation Cards in Full, No Click Needed, November Core Update Volatility Stays Muted as Rollout Nears Its Second Week, and What 9,658 Reddit Threads in Google Reveal About Forum Content and Search.

More from Linkbaits

Turn this playbook into links

Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.

Start free trial