AI Overviews and the Real CTR Data | Growpha
Ahrefs tracked AI Overviews’ effect on the top-ranking Google result from April 2025 to December 2025 and watched the click loss climb from about 34.5 percent to about 58 percent over that stretch. That isn’t a plateau. It’s a trend line that kept moving in one direction for most of the year it was measured, and any SEO reporting that treats AI Overviews as a one-time adjustment rather than an ongoing shift is already behind.
Where the decline concentrates
The drop isn’t even across query types, and that unevenness is the part worth acting on. Informational queries, the definitions, tutorials, comparisons, and how-to searches that trigger AI Overviews most often, show CTR declines reported as high as 61 percent. These are exactly the queries an AI Overview is built to answer directly on the results page, so a user who already has a complete answer has no reason to click through to a source that would just repeat it.
Transactional and branded queries tell a different story. Commercial searches from people close to a purchase trigger AI Overviews less often, and the CTR hit on the ones that do is comparatively modest. Branded and navigational searches barely trigger AI Overviews at all, and where they do, several studies now show branded CTR going up rather than down: a person searching a company name still wants that company’s own page, and an AI summary doesn’t substitute well for that.
Behavioural data backs this up at the user level. Pew Research Center studied 68,879 real Google searches and found people clicked a traditional organic result just 8 percent of the time when an AI Overview appeared on the page, against 15 percent when it didn’t. That gap explains why zero-click behaviour on AI Overview queries now sits between 80 and 83 percent: roughly four out of five people asking one of these questions get an answer and never visit a website at all.
What this means for measurement, not just strategy
Total search impressions have grown roughly 49 percent since AI Overviews rolled out broadly, while click-throughs across the same period fell by around 30 percent. A site’s impressions can climb in Search Console at the same time its real traffic drops, which makes impressions on their own a worse proxy for performance than they were two years ago. Reporting that still leads with impression growth without pairing it against click volume by query type is measuring the wrong thing, and it will keep telling a client or a stakeholder that visibility is improving while the number that pays the bills moves the other way.
The practical response
Doubling down on informational content that already ranks well but now competes with an AI-generated answer for the same query is close to fighting the tide directly, and the data says that fight is being lost gradually rather than won. The queries worth defending harder are the transactional and branded ones, where AI Overviews barely intrude and where a site’s own page is still what the searcher wants to land on. That doesn’t mean abandoning informational content. It means building it with the expectation that a share of its value now comes from being the source an AI Overview cites, rather than the page a human clicks, which is a genuinely different kind of visibility and needs a genuinely different way of tracking whether it’s working.
Building for citation inside AI Overviews and other AI answer engines runs alongside traditional ranking work rather than replacing it: clear, structured, directly answerable content earns citations in these summaries the way it used to earn featured snippets, and a citation without a click still puts a brand name in front of someone even if it never shows up as a session in analytics.
Measuring success against a moving target
The metric that mattered five years ago, organic sessions from non-branded queries, is telling half the story on a growing share of search terms now. A more honest scorecard for 2026 tracks impressions, clicks, and citation appearances separately by query type, because a 61 percent CTR decline on informational terms and a CTR gain on branded ones are two different problems that need two different conversations with a client, not one blended number that hides both.
AI Overviews have cut click-through rates on the top-ranking Google result from roughly 34.5 percent in April 2025 to around 58 percent by December 2025, but the damage isn't distributed evenly. Informational queries lose the most, transactional queries lose less, and branded queries can gain clicks under an AI Overview. That split should reshape both content strategy and reporting: impressions alone are a worse proxy for performance now than they were two years ago, and citation appearances inside AI answers deserve their own line in a scorecard rather than being ignored or blended into click totals.
Key terms used in this article, extracted and defined by AI.
Google's AI-generated summary shown above traditional organic results for certain queries, synthesising information from multiple sources into a direct answer.
A search where the user gets the information they need directly from the results page, without clicking through to any website.
The percentage of people who see a result in the search results and click it, calculated as clicks divided by impressions.
A search that doesn't include a company or product name, typically reflecting broader topic or category research rather than a search for one specific business.
An instance where an AI-generated answer (in an AI Overview or a separate AI assistant) names or links to a specific source as the basis for part of its response.
Cites specific, sourced percentage figures across multiple studies rather than a general "AI is changing search" framing.
Draws on named research (Ahrefs, Pew Research Center) with specific methodology details like sample size, which generic AI Overview commentary usually skips.
Written for someone who needs to explain a real CTR drop to a client or stakeholder, not someone encountering the concept of AI Overviews for the first time.
The measurement-and-reporting angle, treating this as a metrics problem and not just a content problem, is less commonly covered than the "how to rank in AI Overviews" genre.