Published
The numbers, and their spread
The largest published dataset comes from Seer Interactive, which tracked 53 brands, 5.47 million queries and 2.43 billion impressions between January 2025 and February 2026. On queries where an AI Overview appeared, organic click-through rate was 0.61%. On queries without one, it was 1.62%.
That is a compression of roughly 61%. Ahrefs, examining 300,000 keywords, found the top-ranking page lost 58% of its click-through rate when an Overview was present — close enough to corroborate.
A randomised field experiment produced a smaller number. Removing Overviews from queries that would otherwise trigger them raised outbound clicks from 0.38 to 0.61 per search, implying a 38% reduction attributable to the feature. The same experiment found user satisfaction ratings unchanged, which is the finding Google is most likely to cite and the one publishers are least likely to.
The wider context is a market-level shift. SparkToro's analysis put zero-click searches at 68% of Google queries in early 2026, up from 60% in 2024 — a rise of 7.5 percentage points in twenty-four months, the fastest movement in a decade. Overviews now appear on around 58% of queries.
Why the studies disagree
A range from 38% to 61% is wide enough to change a budget, and the reason for it is worth understanding before either number is quoted in a board paper.
Observational studies compare queries that trigger an Overview against queries that do not. Those two populations are not alike. Overviews appear most often on informational questions — definitional, explanatory, the kind that were already being answered in featured snippets and already converted poorly. Some of the measured gap is a property of the questions, not of the feature.
The randomised experiment sidesteps this by removing the Overview from the same queries and observing the same users. That is why its estimate is lower and, methodologically, the more defensible causal number.
The honest summary is that AI Overviews cause a large decline in clicks, and that observational figures above 50% include an element of selection that the experimental design removes.
The losses are not evenly distributed
Aggregated across four research sources, informational and educational content is losing between 34% and 61% of clicks. Branded queries carrying an Overview are gaining.
That asymmetry is the finding with the clearest commercial consequence, and it is routinely lost in coverage that reports a single average. A page explaining a general concept is competing directly with a machine that can explain the concept in the results. A page someone reaches by searching a company's name is not.
Device behaviour compounds it: mobile users are 66% more likely than desktop users to end a search without a click, which matters disproportionately in mobile-first markets.
Cumulatively, Seer put the decline in organic clicks by the fourth quarter of 2025 at 42% against pre-Overview baselines. That is the number a business should assume applies to its informational content unless it has evidence otherwise.
The compensation nobody expected
One finding runs the other way and deserves more attention than it has received. Brands cited inside AI Overviews earned 35% more organic clicks and 91% more paid clicks than brands that were not cited.
The mechanism is straightforward. Appearing in the Overview functions as an endorsement. A user who reads a company's name in the machine's answer is more likely to click its result below, and substantially more likely to click its advertisement.
This reframes the objective. The realistic goal is not to avoid Overviews or to recover the clicks they absorb. It is to be the source the Overview quotes, and to capture the elevated intent of people who have just been told the company is credible.
What to change in reporting
Three adjustments follow, none of which require new tooling.
Segment traffic reporting by query type rather than in aggregate. Informational and branded traffic are now moving in opposite directions, and a blended organic number hides the fact that one is collapsing while the other grows.
Stop judging explanatory content on sessions. Content that exists to establish credibility should be measured on whether the business is named and described accurately when the question is asked, and on what happens to branded search volume over the following quarters.
Protect the queries where clicks still happen. Specific problems, commercial comparisons, local intent and the company's own name remain click-generating, and branded search still accounts for around 44% of Google queries. That is the traffic an AI answer cannot satisfy on the user's behalf, and it is now the most valuable inventory a website owns.
Common questions
How much traffic do AI Overviews actually cost?
Estimates range from 38% to 61% fewer clicks on affected queries. The 38% figure comes from a randomised experiment and is the more reliable causal estimate; higher figures come from observational studies that compare different sets of queries.
Do AI Overviews hurt every kind of page equally?
No. Informational and educational content is losing 34% to 61% of clicks, while branded queries carrying an Overview are gaining clicks. The effect depends heavily on why someone is searching.
Is it better to be cited in an AI Overview or absent from it?
Cited. Brands appearing in AI Overviews earned 35% more organic clicks and 91% more paid clicks than brands that were not cited, because the citation acts as an endorsement.