Your shopping cart is empty!
Zero-click searches are pegged at 22% and 68%, and both numbers hold up

- Estimates of the share of zero-click searches in the US range from 22.4% to 68.01%.
- The gap comes from the definition of a click, the device mix and the user panels, not from someone's arithmetic error.
- Small publishers lose the most: they have shed 60% of their search traffic in two years.
The very same behaviour — a person asks Google, reads the results and leaves — was measured this year at both 22.4% and 68.01% of searches.
PPC Land gathered those figures into one explainer, and it makes the point clear: none of the authors got it wrong, they were simply counting different things.
Why the numbers differ threefold
The share of zero-click searches is not measured inside Google but through panels: a small group of people agree to have their browser behaviour recorded. Such panels are run by Datos (a Semrush subsidiary) and Similarweb, and before them by Jumpshot, which has since closed.
Every search falls into one of five buckets: an organic click, a paid click, a visit to a Google service, a new query, or "nothing happened".
That is where it gets interesting. A click to YouTube or Maps counts as a click for some researchers and as a zero-click exit for others. Add different weightings of desktop and mobile and different panel compositions, and a threefold gap stops being surprising.
I wouldn't ask which of these figures is "right" at all. The questions worth asking are what exactly was counted as a click, and on what sample. Without those two lines, any percentage in a deck is just a big number.
The metric has a long history. Google launched the Knowledge Graph in May 2012 and rich snippets in January 2014, and in August 2019 Rand Fishkin's analysis first showed that more than half of searches end without a click (50.33%). After AI Overviews launched in May 2024 the numbers went up:
| Source | When | What was measured |
|---|---|---|
| Rand Fishkin | August 2019 | 50.33% of searches with no click |
| Datos and SparkToro, desktop | March 2026 | 22.4% |
| SparkToro, US | June 2026 | 68.01% |
| Similarweb, news queries | May 2025 | 69%, against 56% a year earlier |
Who pays for the lost clicks
Ahrefs calculated that AI Overviews cut clicks to top-ranking pages by 34.5% (April 2025) and by 58% (February 2026).
According to Chartbeat, small publishers lost 60% of their search traffic over two years, mid-sized ones 47%, large ones 22%.
The hundred largest media outlets, by the article's estimate, spend around $113 million a month on paid search, 274% more than three years ago. Traffic that used to arrive for free is now partly bought back.
Google does not accept this picture. Back in August 2025 Liz Reid of Google argued that the total volume of organic clicks from Search to websites remains relatively stable year over year, and disputed the methodology of the studies, Pew Research among them.
Regulators are not waiting: the European Commission opened an investigation into the use of publisher content in AI on 9 December 2025, the UK's CMA required Google to allow opting out of AI Overviews without losing visibility (the requirements take effect on 3 December 2026), and Penske Media filed an antitrust suit in the US back in September 2025.
What of this reaches Ukrainian reports
All these figures are about the US, so they cannot be transferred to Ukrainian results directly.
The mechanics, though, are easy to check on your own site. In Search Console, take the queries Google answers outright (exchange rates, weather, "what is…") and compare their impressions and clicks with commercial queries.
If informational queries show plenty of impressions and few clicks, you are looking at your own zero-click without any panels. To tie it to sales it helps to keep SEO reports in GA4 alongside, and how AI answers eat into clicks is covered in detail in our AI Overview piece.
There is a fresh wrinkle too: since 31 August 2026 Google has been rolling out AI impression reports in Search Console, where impressions appear without clicks and without queries. Measuring the effect with your own hands is about to get harder.
One more thing stands out: every party to the argument earns money from its own version of this number. SparkToro sells audience research, Datos and Similarweb sell panels, and Google sells ads in the very results being discussed.


