Key Takeaways
- Pew Research Center tracked 68,000 real search queries and found click-through fell from 15% to 8% when an AI summary was present.
- Ahrefs measured a 34.5% drop in click-through to the first organic result on informational keywords.
- Similarweb put zero-click searches at 69% by May 2025, up from 56% a year earlier.
- Branded queries move the other way: Amsive recorded an 18% CTR increase where AI Overviews appeared on brand terms.
Every few years the search industry has a scare that turns out to be survivable. Panda, mobilegeddon, the featured snippet: each was going to end organic traffic and each ended up reshuffling it. The instinct to treat AI Overviews as the next item on that list is understandable. The data increasingly suggests it does not belong there.
The most useful measurement so far comes from the Pew Research Center, which tracked 68,000 real search queries rather than modelling them. Where an AI summary appeared, click-through to a traditional result ran at 8%. Where it did not, the same behaviour produced 15%. That is close to halving the value of a ranking without changing the ranking at all.
The Numbers Are Consistent Across Very Different Methods
What makes the Pew figure hard to dismiss is that independent measurements using different methods land in the same territory. Ahrefs, in analysis covered by Search Engine Journal, found a 34.5% reduction in click-through to the first organic position on informational keywords. Similarweb tracked zero-click searches rising from 56% to 69% between May 2024 and May 2025.
At the sharp end the declines get much steeper. DMG Media reported desktop click-through on affected queries falling from 25.23% to 2.79%, a drop of 89%. Chegg recorded a 49% decline in non-subscriber traffic across a twelve-month window. Digital Content Next, whose membership is made up of premium publishers, put the median year-over-year decline in Google referral traffic at 10%.
The spread between a 10% median and an 89% worst case is the part worth sitting with. This is not a uniform tax on everyone with a website. It is concentrated almost entirely on one kind of content, and most marketing teams produce a great deal of it.
Informational Content Is the Target
AI Overviews are built to answer questions, so they intercept queries whose entire purpose was to get an answer. The explainer, the definition post, the how-to guide: these were the reliable top of most content funnels, and they are precisely what a summary makes redundant. Transactional and navigational searches are far less affected because the user still has to arrive somewhere to act.
The interface arithmetic reinforces it. Analysis of 8,000 keywords by Advanced Web Ranking found AI Overviews averaging around 169 words with roughly seven links when expanded, pushing the first organic result to approximately 1,674 pixels down the page. Even a reader inclined to click has to travel a long way to find something to click on.
Google's position, put by Liz Reid, its Head of Search, is that queries showing AI Overviews see around a 10% increase in usage and that click volume has remained relatively stable year over year, with click quality improving. Both things can be true at once: more searching, more answers delivered in place, and fewer of those searches ending on a publisher's page.
The Exception That Points at the Strategy
One finding runs directly against the trend and is the most actionable in the set. Amsive measured an 18% increase in click-through for branded queries where AI Overviews appeared, against declines of 34% to 46% on generic terms.
That asymmetry is the whole strategic argument in one statistic. When someone searches for an answer, the summary can satisfy them. When someone searches for you, it cannot, because the thing they want is not information but a specific organisation. Demand that arrives with a name attached is largely insulated from the squeeze.
What Content Teams Should Actually Change
The reflex response is to optimise for citation inside the summary. That is worth doing and it is not a strategy, particularly given that Authoritas found roughly 70% of AI Overview-cited pages changed over a two to three month window, with the changes unlinked to traditional organic rankings.
- Separate your informational and branded reporting. Blended organic numbers will hide a collapse in one and growth in the other until the aggregate finally turns.
- Stop funding explainer content that only answers. If a summary can fully satisfy the query, the page is competing with a free answer placed above it.
- Move budget toward owned channels. Email, SMS and community reach an audience directly rather than renting access to it.
- Write for the searches a summary cannot end. Original data, named practitioner experience and genuine proprietary analysis cannot be synthesised from elsewhere because they do not exist elsewhere.
- Treat brand search as a pipeline metric. It is now one of the few organic signals reliably moving in the right direction.
- Expect citation volatility. With most cited pages turning over within a quarter, visibility inside an Overview is not a position you hold, it is one you keep re-winning.
The uncomfortable conclusion is that a decade of content strategy was built on an arrangement that no longer holds: publish the answer, earn the ranking, collect the visit. Two of those three steps still work. The third has been quietly removed, and no amount of optimisation puts it back.
