Do AI Overviews kill clicks? Data for small sites
Every major study says Google's AI answers cut clicks on the queries where they appear, and Google says clicks are stable. Both are measuring something real. Here is the full data set, what it means for a small business site, and the one finding that changes the strategy: citation flips the effect.
When Google started answering questions directly at the top of its results, every site owner asked the same thing: is this eating my clicks? Three years of studies later, the answer is documented from several independent directions, and it's more specific than the doom headlines or Google's reassurances suggest.
The short version: yes, an AI Overview substantially reduces clicks on the query where it appears. But the damage is concentrated almost entirely in informational queries, the effect partially reverses if the AI answer cites you, and the queries closest to money remain the least affected. For a small business, that combination points to a clear strategy rather than despair. Here's the data behind each piece.
How often you're even competing with an AI answer
First, scale. Two large 2025-2026 measurements landed on nearly the same number: Conductor's benchmark found 25.11% of 21.9 million US queries triggered an AI Overview, and SEO.com's study of 2.37 million queries found 25.8%. Roughly a quarter of searches now open with an AI answer.
The distribution matters more than the average. SEO.com found 38.7% of informational queries showed an AI Overview, rising to 67.2% for long, question-shaped searches of eight or more words, while queries with buying intent triggered them far less. Ahrefs' data agrees from another angle: in its 300,000-keyword study, 99.2% of the keywords with AI Overviews were informational. The AI answer lives where people ask questions, not, so far, where they pull out wallets. That gap is closing though: Semrush's tracking of 600,000 US keywords found commercial-intent queries triggering AI Overviews grew 71% in the six months to April 2026. The quarter-of-searches figure is a floor, not a ceiling.
What happens to clicks when the answer appears
The most trustworthy single study is Pew Research Center's, because it measured real people rather than dashboards: 900 US adults with browsing trackers, 68,879 real Google searches in March 2025. When a search showed an AI summary, users clicked a regular result 8% of the time, versus 15% without one. The click rate on the queries with AI answers roughly halved. Users also ended their session entirely more often with a summary present, 26% versus 16%, and clicks on the sources cited inside the summary were nearly nonexistent, happening in about 1% of visits.
The tool-data studies show the same shape over time, and show it getting steeper. Ahrefs compared 150,000 AI Overview keywords against 150,000 without, and found the top-ranking page's click-through rate came in 34.5% below forecast once AI Overviews rolled out. By its February 2026 update, the position-one click rate on those queries had fallen to roughly a fifth of what it was two years earlier. Seer Interactive, tracking 3,119 informational terms across 42 client organizations, watched organic click-through on those terms fall from 1.76% in mid-2024 to 0.61% by September 2025.
Different methods, same conclusion: on question-type queries where the AI answers first, most of the classic click opportunity is gone.
What Google says, and how both things are true
Google's public position, stated by Search head Liz Reid in August 2025, is that total organic click volume is "relatively stable year-over-year" and that "average click quality has increased and we're actually sending slightly more quality clicks to websites than a year ago", defining quality clicks as ones where the user doesn't immediately bounce back.
Site owners tend to hear that as denial, but the two data sets don't actually contradict. Google is describing totals across all search, which keeps growing in query volume. The independent studies describe rates on the specific queries where AI Overviews appear. Clicks per informational query can crater while total clicks stay flat, especially if people search more because answers are easier to get. And the "quality" framing contains a real insight for you: the clicks that survive an AI answer come from people who wanted more than the summary, which makes them better visitors on average. Fewer, warmer clicks is a real trade, not spin, but it is a trade someone else decided for you.
The finding that changes the playbook: citation flips the sign
Buried in Seer's data is the strategic core of this whole topic. Brands cited inside the AI Overview had a 35% higher organic click rate on those queries than brands that weren't, and paid ads alongside a citation performed 91% better. By its 2026 update, covering 2.43 billion impressions, pages cited in the AI Overview were earning roughly 120% more organic clicks per impression than uncited pages. The same update caught the first recovery signal in the series: organic click rates on AI Overview queries climbing from 1.3% in December 2025 to 2.4% in February 2026, as citation rates rose and users apparently learned to click through the answers.
Read those together: the AI Overview doesn't just take clicks, it redistributes them, away from the classic blue links and toward the handful of sources the answer names. Being invisible on an AI-answered query now costs more than losing a rank ever did, and being cited recovers a disproportionate share of what's left. Which is why earning citations in AI Overviews has become its own discipline, and why the extractable, source-worthy content it requires overlaps almost perfectly with what the other AI search engines reward.
What this means for a small business site
Translating the data into decisions:
- Reweight, don't abandon, informational content. Its job has changed. A how-to post now buys visibility, brand impressions inside AI answers, and topical depth more than it buys direct clicks, the dynamic zero-click search already established before AI Overviews accelerated it. Publish informational content that demonstrates expertise and feeds citations, and stop measuring it by clicks alone.
- Point your click expectations at commercial queries. Comparison pages, service pages, pricing explainers, local-intent pages: these still trigger AI answers least and still deliver classic clicks, and they're the queries buyers use anyway. The 71% growth in commercial AI Overviews says this window narrows, which argues for building those assets now, not later.
- Compete for the citation, not just the rank. Clear, self-contained answers, real data, named expertise: the content traits that get lifted into AI answers. Your ranking still matters, since cited sources overwhelmingly come from pages Google already trusts, but the citation is the new position one on informational queries.
- Watch your own numbers, not the averages. Every figure above is an average across industries. In Search Console, flat-or-rising impressions with falling clicks on your question-type pages is the AI Overview signature, and that diagnosis tells you which of your pages are affected and which citation opportunities to chase first.
Measuring the effect on your own site
The studies give you the market average; your exposure is knowable more precisely, from data you already have. The method takes half an hour in Search Console.
Start with the queries report, comparing the last three months against the same three months a year ago. Sort by impressions and look for the signature pattern: queries where impressions held steady or grew while clicks fell and average position didn't materially change. Stable position plus stable impressions plus falling clicks means the demand still exists and your ranking still exists, but something between the ranking and the click is absorbing attention, and on question-type queries in 2026, that something is usually an AI Overview. Spot-check by searching a few of the flagged queries yourself: if an AI answer sits on top, you've confirmed the mechanism.
Then split your flagged queries into two piles by intent. The informational pile, how, what, why questions, is where the studies predict the damage concentrates, and where your remaining options are earning the citation or accepting the visibility-not-clicks trade. The commercial pile, anything with cost, best, near me, vs, or a service name, deserves closer attention: these queries still deliver classic clicks, and any erosion here matters more per click lost. If your commercial queries are still clean of AI Overviews, that's your window: the Semrush growth figure says commercial coverage is expanding, so the defensive work, stronger comparison pages, extractable answers, citation-worthy specifics, is best done before the answer box arrives, not after.
One trap to avoid while diagnosing: a site-wide click decline has many possible causes, rankings, seasonality, a technical problem, and blaming AI Overviews without the per-query signature is how sites misdiagnose ordinary SEO problems as acts of Google. The signature is specific, per-query, and checkable. Diagnose at that resolution and your response can be specific too.
Quick answers to the follow-ups
Can I opt my content out of AI Overviews? Not selectively in any practical way: the controls that keep content out of AI features overlap with the ones that limit ordinary snippets, so opting out trades away regular visibility too. For most sites the real choice is between competing for citations and ignoring them, not between appearing and not appearing.
Do AI Overviews show for local searches? Far less often than for how-and-why questions, and local queries retain their map results, which remain the local battleground that matters. A local service business's click exposure is concentrated in its informational blog content, not its core "service + suburb" queries, which is reassuring for revenue and clarifying for strategy.
Is this the same thing as zero-click search? It's the acceleration of it. Zero-click results predate AI, featured snippets and answer boxes started the pattern years ago, but AI Overviews extend it to longer, more complex questions that previously required a visit to answer.
Should I stop writing informational content entirely? No, reprice it. It still builds the topical depth that rankings and citations rest on, still seeds the brand impressions that turn into branded searches, and still feeds the AI answers themselves. What it no longer does reliably is deliver the click, so stop buying it for clicks.
The honest bottom line
On the queries where they appear, AI Overviews reduce clicks, roughly by half in the cleanest user-level study, and the footprint is spreading from questions toward commercial searches. That's the loss, and it's real. The offsets are also real: a quarter of queries affected is not all queries, cited sources claw back an outsized share of the remaining clicks, and the visitors who do click are more committed than the drive-by traffic that's disappearing.
For a small site the conclusion isn't to quit search, it's that the deliverable changed. You're no longer only ranking pages for clicks. You're building the kind of specific, trustworthy, extractable content that gets your name into the answer, and collecting fewer, better visitors on the queries that lead to revenue. Search isn't dying, but the informational click subsidy that funded a decade of blogging is being repriced, and the sites that adapt to citation economics first will take the traffic the others lose.