Does Google AI Overview Hurt Your Website Traffic?

Four large studies have measured this since 2025. They disagree about the size of the effect, agree on its direction, and each carries a limitation its headline number omits. Here is what each actually measured, before what it means for one specific site.

Read the middle column and the pattern is visible. The question is not whether AI Overviews reduce clicks, but which of your pages sit in the blast radius, and that set is narrower than the panic suggests.

Which queries actually lose clicks, and which hold up?

Exposure is not evenly distributed, and the distribution is the whole story. Seer Interactive's April 2026 analysis of 5.47 million queries found AI Overviews on 36% of informational searches, 8% of commercial ones and 5% of transactional ones. A page cannot lose clicks to a feature that never appears above it. If your traffic comes from people typing a service and a city, or a product name and the word buy, most of your impressions never carried an AI Overview at all.

The local numbers are where careless reporting does the most damage. Whitespark's 2026 study is widely quoted for the line that AI Overviews now appear on 68% of local business queries, which sounds like an emergency. Decompose it and the emergency dissolves. That 68% is a blended average across three query categories the researchers deliberately separated: local intent at 15%, informational intent at 92%, and hybrid intent at 97%. The blended figure is arithmetically true and practically misleading. A dentist ranking for "dentist Brantford" sits in the 15% bucket. The same dentist's article on what a root canal involves sits in the 92% bucket. Those are two different problems.

We covered the mechanics of how these panels differ from the older snippet format in our research note on AI Overviews versus featured snippets, and the click-behaviour side specifically for service businesses in zero-click search and the local business.

How big is the click loss when an AI Overview does appear?

Large, on informational queries, by every independent measurement so far. Pew Research Center tracked the real browsing of 900 US adults through KnowledgePanel Digital during March 2025, covering 68,879 Google searches, 12,593 of which produced an AI summary. On results pages carrying a summary, users clicked a traditional result link on 8% of visits. Without one, 15%. The share ending their session entirely rose from 16% to 26%.

One Pew figure deserves more attention than it gets: users clicked a link inside the AI summary itself on 1% of visits. The citation links in an AI Overview are not a meaningful traffic channel in their own right. They are a visibility and attribution signal whose value shows up elsewhere, which is a different proposition and worth being precise about.

Ahrefs measured the same phenomenon through aggregated Search Console data across 300,000 keywords, finding that an AI Overview correlated with a 34.5% lower click-through rate for the top-ranking page in April 2025. Re-running the identical method against December 2025 data, published February 2026, the correlation had widened to 58%. Both numbers circulate today as competing claims about one fact. They are the same measurement at two points in time, and either should be cited with its date attached, because the trend between them says more than either endpoint does.

Seer Interactive's figures put a floor under all of this and add the detail that changes what you do about it. Across informational queries, average organic click-through rate ran 3.35% where no AI Overview appeared, 2.07% where an AI Overview appeared and the brand was cited in it, and 0.94% where an AI Overview appeared and the brand was not cited. Citation roughly doubles the click-through rate among AI Overview queries. It does not restore the pre-Overview number.

Why does Google say traffic is stable when the studies say it fell?

Because the two sides are measuring different quantities, and only one side has published its working. On 6 August 2025, Liz Reid, VP and Head of Google Search, wrote that "total organic click volume from Google Search to websites has been relatively stable year-over-year" and that "average click quality has increased." Google defines a quality click as one "where users don't quickly click back," treating a fast return to the results page as a signal the visit was not useful.

Take the claim at face value and it stays compatible with everything above. Total click volume is an aggregate across the entire web, on a search engine whose query volume Google says is growing. A stable total is consistent with a large redistribution underneath it: fewer clicks to definitional content, more to commercial and local pages, and a reshuffling of who receives them. Aggregate stability and per-site collapse are not contradictory. Most publishers reading the reassurance are not asking about the aggregate.

What should temper the weight anyone gives the claim is procedural. Google's post contains no figures. No baseline, no percentage change, no time window, no methodology, no sample definition. It criticises third-party studies for flawed methodology while disclosing none of its own. That is not evidence of dishonesty and we will not imply it is. It does mean the claim cannot be checked, and an unverifiable claim from an interested party is worth less than a published methodology with stated limitations.

What can nobody outside Google actually verify?

The specific click count. This is the honest centre of the question and it is routinely skipped by pages selling a solution to it.

Google launched Search generative AI performance reports in Search Console in June 2026. They show impressions, pages, countries, devices and dates for URLs appearing in generative AI features. They do not show clicks. No click count, no click-through rate, no average position, no query data for AI Overviews specifically. Google's documentation states that sites appearing in AI features "are included in the overall search traffic in Search Console," inside the Web search type alongside everything else.

Ahrefs said this plainly in their own study, and it is unusual enough for a vendor to flag a limitation that undercuts their headline that it deserves quoting: there is still no way to disambiguate AI Overview clicks and impressions from the rest of your Search Console data.

So when a tool or an agency gives you a precise figure for clicks lost to AI Overviews, that number was modelled, not measured. Modelling is legitimate and some of these models are careful. But the line between a measured number and an inferred one is exactly the line this niche keeps erasing, and once erased you cannot tell a good estimate from a confident guess.

How do you check your own Search Console for the pattern?

You cannot isolate AI Overview clicks, but you can detect their fingerprint, because the click compression they cause leaves a distinctive combination of movements that ordinary ranking loss does not produce. This takes about twenty minutes in a free tool you already have.

If that check surfaces something you want a second opinion on, our audit runs the same segmentation against your account and reports what it finds.

What do the four possible results mean?

Four combinations are possible, each pointing somewhere different. Only one is an AI Overview story.

That last row causes more false alarms than AI Overviews do. In an anonymised Ontario shipping container dealer engagement, a 16-month window recorded 4,810 clicks against 627,000 impressions, with single-digit daily clicks through most of the first year before seventy-click days arrived by early summer 2026. Blended average position sat at 26.3, a number that reads as failure until you see it is averaging established commercial pages near the top against a growing tail of new pages entering far lower. The blended figure worsened while the business improved. Averages across a changing page set are among the easiest ways to misdiagnose your own data (Google Search Console, 16-month window, shared with permission).

Is the protected zone shrinking?

At the commercial edge, yes, and anyone planning around this should treat the boundary as moving. Semrush analysed more than 600,000 US desktop keywords between November 2025 and April 2026 and found the share of commercial-intent results pages carrying an AI Overview grew 71% over those six months. Finance was the extreme at 231%, computers and electronics 107%.

The same study found the share of transactional results pages carrying an AI Overview fell 5% across the identical window. Google appears to be extending these panels through the research and comparison phase while leaving the point of purchase largely alone, which is a coherent product decision: a summary helps someone deciding between options and gets in the way of someone ready to act.

The consequence is that the safe list from eighteen months ago is not the safe list today. Comparison pages, "best X for Y" pages and buying guides have moved toward the high-exposure column. Treat query-class exposure as something to re-measure quarterly rather than a fact you learned once.

What actually changes the outcome?

Two things, one of which is unglamorous and mostly involves deciding what not to build.

The first is citation inside the panel where it appears. Seer's spread between 2.07% cited and 0.94% uncited on the same informational queries is the clearest quantified argument for being named in the summary rather than merely ranking beneath it. Google's own AI features documentation is unusually direct about how that is earned, and it contradicts much of what is currently sold: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." Google describes query fan-out, where its models break a query into sub-questions and assemble supporting pages for each, which is why pages answering one specific question completely get pulled in more often than pages surveying a topic broadly.

The second is portfolio composition. If much of your organic traffic depends on definitional content an AI summary can fully satisfy, that exposure is structural and no amount of on-page work removes it. The response is to shift weight toward pages a summary cannot replace: your pricing, your process, your service area, your inventory, the specifics of how you do the work. Those are also the pages that convert, which makes this less a crisis than a forced reallocation toward content that was always worth more per visit. The quality system behind these calls is the one we broke down in our note on the helpful content system and AI.

The verdict

Does Google AI Overview hurt your website traffic? For informational and definitional pages, the independent evidence says yes, consistently: Pew measured 8% versus 15% click rates in real browsing, Ahrefs a correlation widening from 34.5% to 58% in under a year, Seer click-through rate falling from 3.35% to 0.94% when a brand appears on an AI Overview query uncited. For transactional and genuine local-intent pages, exposure stays low enough that most such sites are diagnosing a different problem.

What nobody outside Google can tell you is the exact figure for your site, because that click data does not exist in any tool available to you. Google says the aggregate is stable and has published nothing to support it. The independent studies say the per-query effect is large and have published methodologies you can read and argue with. Both positions survive contact with the other, and treating either as settled is the mistake. What you can establish in an afternoon, from data you already own, is which of the four patterns your account is showing. Start there. We publish more work like this in our research library.