The query is typed
Nothing has changed here. Search volume did not fall. People still ask, at the same rate, in the same language.
Search, 2026
The click that never came
Google spent twenty years sending people to other people's websites. In 2026 it mostly stopped. Here is the shape of what happened, and why the averages hide the real story.
A search used to end somewhere else
For most of the web's commercial life the bargain was simple. You asked Google a question, Google showed you ten places that might answer it, and you went to one of them. The visit was the payment. Everything downstream of that visit, the advertising, the affiliate links, the email signups, the small businesses that lived off a first page ranking, all of it rested on the click actually happening.
AI Overviews changed the last step. The answer now appears above the results, in Google's own words, assembled from those same ten places. The question still gets asked. The question still gets answered. The visit is the part that goes missing.
39.8%
34.5%
58%
Two studies, two numbers, one direction
The peer reviewed figure and the Ahrefs figure are not the same measurement and should not be averaged into one. The first asks what happens to clicks on a query when an overview appears versus when it does not. The second asks what happened to the click through rate of pages that rank at the very top, which is exactly the position an overview now sits above.
They disagree by nearly twenty percentage points and they agree on everything that matters. The effect is large, it is not a blip, and it lands hardest on the results that used to be worth the most.
Sixty eight percent of Google searches in the first four months of 2026 ended with no click at all.
The number that reframes everything else
That is the fastest acceleration of the zero click trend in a decade. It is worth sitting with, because it inverts an assumption almost every website was built on: that a search is the beginning of a journey. For two searches in three, the search is now the whole journey.
How one search stopped being a visit
- 01The query is typed
- 02The overview assembles
- 03The answer satisfies
- 04The links go unread
- 05The publisher sees a flat line
The overview assembles
Google reads the pages that would have ranked, and writes a short answer in its own voice. The sources are cited, in a collapsed panel, below the answer.
The answer satisfies
For a factual question, this is where it ends. The reader has what they came for and no reason to continue.
The links go unread
The ten blue links are still there. They are now below an answer, and below the fold on a phone.
The publisher sees a flat line
Impressions hold up. Clicks do not. This is why so many sites noticed the revenue before they noticed the traffic.
The averages are hiding the story
Reported publisher losses run from one percent to over seventy five percent. That spread is not noise, and it is not a measurement problem. It is the finding.
This is not a uniform tax applied evenly to everyone with a website. It is close to a total wipeout for one kind of page and barely a scratch for another. Which kind you are matters far more than what the average says.
The same subject, two fates
Loses the click
A page that answers "how tall is a blue whale" in a paragraph, surrounded by advertising, ranking first for a question with a one line answer.
The overview says thirty metres. The reader nods and closes the tab. Nothing about the page was bad. It was simply a slower way to get the same sentence.
Keeps it
A page that puts a blue whale on the screen at true scale and makes you scroll its entire length to reach the tail.
The overview can still say thirty metres. It cannot do the thing the page does. The answer and the experience have come apart, and only one of them is substitutable.
What still works
Questions a sentence cannot close
Comparison at scale, a process you have to walk through, change over time, anything where the answer is a shape rather than a fact.
Being the source, not the summary
Overviews cite. Cited sources keep a residual click. Owning original data is the way onto that list; aggregating other people's is the way off it.
Traffic that was never rented
Email, direct, embeds, word of mouth. Narrow channels, but nobody can adjust them out from under you.
Pages that earn something other than impressions
If a page's job is to produce enquiries rather than ad views, the traffic it needs to be worth running drops by about two orders of magnitude.
The uncomfortable part
There is a comforting version of this story where the sites that lost traffic deserved to, and the good ones will be fine. It is not quite true.
An overview does not need to be better than your page to cost you the visit. It only needs to arrive first and be sufficient. A great page about a simple question loses to a mediocre sentence about the same question, because the reader was never grading the answer. They were finishing a task.
Sources
Every figure on this page comes from published research, listed below. Where we have estimated something, the page says so in the place where the estimate appears.
Sources
- Researchers find Google AI Overviews cut publisher clicks 39.8 percent: https://ppc.land/researchers-find-google-ai-overviews-cut-publisher-clicks-39-8/
- Impact of AI Overviews, and how publishers need to adapt: https://www.searchenginejournal.com/impact-of-ai-overviews-how-publishers-need-to-adapt/556843/
- Google AI Overviews, 2026 updates: https://www.seo-kreativ.de/en/blog/google-ai-overviews-updates-2026-en/