Ask what AI Overviews are doing to web traffic and you get two answers that cannot both be casually true. Google's is calm: total organic click volume is "relatively stable" year over year, and AI Overviews are actually sending "slightly more quality clicks" than before, because people search more and see more links. The publisher-side answer is not calm at all. Chartbeat measured Google organic search traffic to more than 2,500 sites down about a third globally in the year to November 2025; Pew found people click a traditional result just 8% of the time when an AI summary appears, versus 15% when it does not; Penske Media, owner of Rolling Stone and Variety, sued Google in September 2025 over exactly this. The instinct is to pick a side. The more useful move is to see how both are partly true at once — Google's aggregate can hold steady while the distribution underneath it shifts hard, concentrating the surviving clicks on a minority of cited brands and draining them from everyone else. That reconciliation is the whole story, and it changes what you should measure. This guide lays out Google's exact claims, the independent data against them, the mechanism that lets both be true, why the traffic you can measure now misses a growing share of the encounters you are actually getting, and how to read your own site's reality instead of either headline.
Ask what Google's AI Overviews are doing to web traffic and you get two answers that seem to flatly contradict each other. Google's answer is calm. In an August 2025 post, its head of Search said total organic click volume from Search to websites has been "relatively stable year-over-year," that it is sending "slightly more quality clicks" than a year ago, and that more searches and more visible links add up to more chances for sites to get found. The publisher-side answer is not calm. Chartbeat measured Google organic search traffic to more than 2,500 sites down roughly a third globally in the year to November 2025; Pew found people click a traditional result just 8% of the time when an AI summary appears, versus 15% when it does not; and Penske Media, which owns Rolling Stone and Variety, sued Google in September 2025 over the traffic loss it attributes to AI Overviews.
The reflex is to decide which side is lying. Neither is. The more useful move — and the one this guide is built around — is to understand how both can be substantially true at the same time, because that reconciliation is what actually tells you what to do. Google reports the aggregate across the whole web, and an aggregate can hold flat while the distribution underneath it shifts hard: some sites up, many sites down, the surviving clicks pooling on a minority of cited brands. Once you see that, the argument stops being "is traffic falling or not" and becomes the two questions that matter for your site specifically: are you on the winning or losing side of the redistribution, and how would you even know? The second question is the harder one, because AI Overviews have changed not just how much traffic you get but how much of your actual visibility is measurable at all.
Take Google's claims exactly as stated, because they are more careful than the paraphrases. Writing on August 6, 2025, Liz Reid, the VP who heads Search, made four points. Total organic click volume from Google to websites is "relatively stable year-over-year." Average click quality has risen, so Google is sending "slightly more quality clicks" than a year ago — where a quality click means one the user does not quickly bounce back from, read as a signal of genuine interest. People are searching more, with longer and more complex queries, and they are seeing more links on the page, so there are "more opportunities for websites to surface and get clicked." And Google continues to "send billions of clicks to websites every day." Reid explicitly rejected third-party reports of "dramatic declines," attributing them to flawed methodology or data predating the AI rollout.
Two things about that position are worth holding onto. First, the "quality clicks" argument has a real mechanism behind it: AI Overviews can answer the shallow, click-and-bounce informational queries directly, so the clicks that do reach your page are, on average, from people who wanted more than a quick fact. That is not spin; it is plausibly what happens to the composition of clicks when the low-value ones are absorbed by the answer box. Second, and more important, Reid conceded the exact thing that dissolves the apparent contradiction: "user trends are shifting traffic to different sites, resulting in decreased traffic to some sites and increased traffic to others." That sentence is easy to skim past, but it is the whole game. Google is telling you the total is stable and the distribution is moving. If your site is one of the ones the distribution is moving away from, Google's aggregate being flat is no comfort at all — and it is not evidence that your decline is imaginary.
Against the aggregate sit a stack of independent measurements that are hard to wave off, because they come from different methods and land in the same place. Pew Research is the cleanest: it recruited over 900 U.S. adults who installed a browsing tracker, then analyzed 68,879 real Google searches from March 2025. People clicked a traditional search result 8% of the time when an AI Overview appeared, versus 15% when one did not — roughly half — and clicked a link inside the summary itself only about 1% of the time. Chartbeat, aggregating over 2,500 sites, found Google organic search traffic down about 33% globally and 38% in the United States between November 2024 and November 2025. Digital Content Next, the trade body whose members include the New York Times and Condé Nast, reported a median 10% year-over-year drop in Google search referrals across 19 members in mid-2025. Seer Interactive, tracking millions of queries, measured organic click-through rate on AI Overview queries falling from 1.76% to 0.61% year over year — a roughly 61% drop.
Then there is the legal signal, which matters because a lawsuit puts a company's own internal numbers on the line. Penske Media Corporation — Rolling Stone, Billboard, Variety, The Hollywood Reporter, Deadline — filed a federal antitrust suit against Google on September 12, 2025, amended in December, arguing that AI Overviews lift its journalists' work, answer the query on the results page, and keep readers from clicking through, all while Penske has no real way to opt out without disappearing from Search entirely. Google has moved to dismiss, calling AI Overviews a product improvement, but in 2026 hearings the judge pushed back that the arrangement "seems really unfair." You do not sue your largest referral source, and stake your internal traffic data on it, over a decline you invented. The publisher experience is real, it is widespread, and the deeper aggregate picture is covered in the publisher traffic collapse guide.
Here is the reconciliation, and it is not a fudge. Google reports one number — total clicks across the entire indexed web — and that number can be genuinely stable while almost every individual publisher you talk to is genuinely down. The way an aggregate stays flat while its members fall is redistribution: for every site losing clicks, clicks are accruing somewhere else. The question is where, and the answer is the mechanism that makes AI Overviews a winners-and-losers machine rather than a uniform tax. Seer Interactive's analysis found that brands cited inside an AI Overview earn dramatically more organic clicks per impression than uncited brands on the same query — on the order of double. So the clicks that survive the answer box do not vanish; they concentrate on the handful of sources the Overview actually names and links, and they drain from the majority it summarizes without citing.
Stack that on top of the raw click compression and the picture resolves cleanly. AI Overviews reduce the total clicks a given informational query produces — that is the Pew and Ahrefs and Seer finding, worked through in AI Overviews are reducing organic clicks. But of the clicks that remain, a larger share funnels to cited brands, big trusted publishers, and destinations users specifically seek out. So a small or mid-size site sees a steep drop (fewer total clicks, and it is rarely the one cited), a large cited brand may see clicks hold or even rise (fewer total, but a much bigger slice), and Google's sum across all of them looks "relatively stable." Google's "quality clicks" framing describes the same thing from the composition angle: the surviving clicks skew toward deeper engagement because the shallow ones were absorbed and because they land more often on sources the user was steered toward. Every one of these statements can be true simultaneously. Whether it is good news for you depends entirely on which side of the redistribution your site sits — which is precisely what the aggregate cannot tell you and what you have to measure yourself.
There is a subtler shift underneath the click numbers, and it is the reason "is my traffic down" has become a bad proxy for "is my visibility down." When an AI Overview names your brand but the reader does not click — the most common outcome, since Pew found source-link clicks at about 1% — you have still been seen. The reader learned you exist, what you do, and whether you were recommended. That encounter shaped an impression and sometimes a decision, and none of it registers in a single line of your analytics, because analytics counts visits and this was a visibility event with no visit. Google even folds AI Overview appearances into Search Console impressions, so a page cited inside an Overview can rack up impressions it earns almost no clicks from, which is exactly why the click-through-rate line can fall while impressions hold. The interface changed; your ranking did not.
This is what people mean when they say AI Overviews reshaped how search visibility is measured. The traffic report was never a measure of visibility directly — it was a measure of clicks, which used to be a reliable stand-in for "people encountered and considered us." AI Overviews broke the stand-in. A page can be losing clicks while gaining influence: cited more often, mentioned to more buyers, driving more branded searches later, and the click chart reads as pure decline. Read your traffic number in isolation now and you will systematically misjudge your own performance — either panicking over a decline that is structural and partly offset by uncounted awareness, or missing that you have genuinely fallen off the surfaces that matter. The case for why click-based measurement stopped working, and what replaces it, is made in full in AI content didn't stop working — your metrics did and AI search and SEO KPIs.
Because the aggregate cannot answer the only question that matters for you, the practical work is reading your own site — and the pattern has a distinct signature. In Google Search Console, an ordinary ranking problem shows average position getting worse with clicks and impressions falling together. The AI Overview pattern is different and specific: impressions holding steady or rising, average position flat or improving, and clicks and click-through rate falling anyway. That combination means Google is still surfacing your page as prominently as ever, but the click is being absorbed by the answer above it. When position is stable and CTR is sliding, the interface changed, not your ranking — and no amount of extra SEO will fix a click Google has decided to keep. The step-by-step version of this diagnosis is in how to measure traffic lost to AI Overviews.
Then segment, because the loss is not uniform and an average hides it. Filter your Performance report to informational, question-shaped queries — "what is," "how to," definitions, comparisons — and compare a recent window against a pre-mid-2024 baseline; that is where AI Overviews trigger most and where the compression concentrates. Contrast it against your brand and transactional queries, which barely move because a user seeking a specific destination or ready to buy still has to leave the results page. Read impressions and clicks together rather than separately, use Search Console's newer AI Overviews and AI Mode impression data to see when you are being surfaced inside an answer (covered in AI search impressions in Google), and pair the whole picture with branded-search volume and direct traffic — a rising branded-search trend not explained by a campaign is one of the cleanest available proxies for AI-driven awareness that produced no click. The distribution and channel consequences of all this are worked through in SEO vs AI Overviews traffic decline.
The redistribution has a shape you can plan against. The losers are concentrated in one place: sites whose traffic was built on ranking for informational queries a paragraph can answer, and that are rarely the source an Overview cites. That is most blogs, most affiliate and how-to content, and most mid-size publishers — the exact model that treated a top ranking on a question as a traffic guarantee. The defended are brand and navigational queries (someone searching your name to reach you, which an answer box cannot substitute for) and high-intent commercial queries where the decision still requires leaving the page. The winners are the cited sources: the brands the Overview names and links, which pull an outsized share of the shrinking click pool. What the winners tend to share is not a trick — it is being a consistently present, well-structured, trusted source that an answer engine is confident naming, across enough surfaces that the model keeps encountering the same claim from the same brand.
That points at the only two responses that actually track the data, and they are complementary, not alternatives. One: work to be a cited winner rather than an uncited loser, because the clicks concentrate there — which is less about a schema checklist and more about being genuinely present and corroborated across the sources AI answers pull from. Two, and more durable: stop betting your distribution on the Google click at all, because even the winners get fewer raw clicks per query than they used to. The reader who encountered you inside an Overview and did not click is still reachable — on the feed where they spend attention, in the inbox they opted into, on the video surface answer engines increasingly cite. The move is to put the same answer in front of people on surfaces where a click to your site is not the required step, so the visibility you are already earning inside AI answers converts somewhere that does not depend on Google forwarding a click it has decided to keep.
Be exact about the boundary. Kompozy is not a traffic analytics tool — it does not read your Search Console, compute your decline, or track your citations, and the dashboards named above do that job. What Kompozy addresses is the production side of the two responses the data actually supports. The first is being a cited winner: the clicks that survive AI Overviews concentrate on brands an answer engine keeps encountering, consistently described, across the blog posts, social feeds, and video it pulls from — corroboration at volume, not a single optimized page. The second is routing around the click entirely: reaching the reader on surfaces where a visit to your site is not the step Google gets to withhold. Both are production problems before they are strategy problems, which is where a normal-sized team stalls.
Kompozy is a full AI content generation and multi-platform publishing engine — 18 output formats across the eight social platforms plus blog and email — driven by one Persona Brief that fixes your voice, claims, and positioning so every piece says the same thing wherever it lands. From a single source it generates the structured, citable blog article built to be the passage an AI Overview quotes, plus the same message restated as text posts, Quote Graphics, Carousel Posts, and a talking-head Persona Short for the feeds and video surfaces those answers increasingly cite — with HyperFrames keeping every asset brand-exact. That is the cross-surface consistency the citation concentration rewards, and simultaneously the feed-native presence that reaches the reader without needing the click, produced in one pass instead of five separate workflows.
Then Autopilot schedules and publishes the whole spread across the supported platforms plus blog and email from one queue, behind a per-post review gate so a person signs off before anything ships — and it generates an email newsletter from the same brief, the one channel no AI Overview, algorithm, or answer box can sit in front of. The honest scope: Kompozy cannot make Google send you a click it has decided to keep, cannot force an Overview to cite you, and will not tell you whether your traffic is up or down — no content tool can. What it removes is the volume ceiling that otherwise makes the two proven responses impossible for a small team to run at once. When AI Overviews have repriced your informational pages and the aggregate Google reports is no comfort because you are on the losing side of the redistribution, the way out is to be present and consistent everywhere the answer is being consumed. Kompozy is what makes that affordable to actually do.
Google and its critics are both telling a version of the truth about AI Overviews and web traffic. Google's total click volume can be "relatively stable" while Chartbeat sees publisher search traffic down a third and Pew sees clicks halve when an AI summary appears, because a stable aggregate is hiding a hard redistribution underneath — the surviving clicks concentrating on cited brands and draining from everyone else. That means the industry-level debate is the wrong altitude for your decisions. The only number that answers your question is your own, read correctly: watch for clicks and CTR falling while position and impressions hold, segment by query intent, and pair it with branded search and AI-surface impressions so you can see the awareness that no longer produces a click. Then act on what the data rewards — be a source AI answers cite, and reduce your dependence on the Google click by meeting people on the surfaces they already occupy. The stable-versus-collapsing argument will run for years. Your traffic reality is knowable now, and it is the only one worth planning around.
For most publishers, yes — but the size depends on who you are. Independent data is consistent: Pew found people clicked a traditional result 8% of the time when an AI summary appeared versus 15% without one, and Chartbeat measured Google organic search traffic to over 2,500 sites down about a third globally in the year to November 2025. Google counters that total click volume is "relatively stable" and that it sends "slightly more quality clicks" than a year ago. Both hold because AI Overviews concentrate the surviving clicks on a minority of cited brands while draining them from everyone else, so the aggregate can look flat while individual sites fall sharply.
In an August 6, 2025 post, Google's Liz Reid stated that "total organic click volume from Google Search to websites has been relatively stable year-over-year," that "average click quality has increased and we're actually sending slightly more quality clicks to websites than a year ago" (quality clicks meaning ones where users don't quickly click back), and that more, longer queries plus more visible links mean "more opportunities for websites to surface." Google explicitly disputes third-party reports of "dramatic declines," attributing them to flawed methodology. Reid also concedes that traffic is shifting between sites — decreasing for some, increasing for others — which is the concession that reconciles the two narratives.
Because they are measuring different things. Google reports the aggregate — total clicks across the entire web — which can stay flat while the distribution beneath it changes dramatically. Publishers report their own sites, and many of them are on the losing side of that redistribution. The reconciling mechanism is concentration: Seer Interactive found brands cited inside AI Overviews earn far more clicks per impression than uncited brands on the same queries, so the clicks that still happen pool among the cited winners. A stable total made of rising winners and falling losers is exactly what both sides are honestly reporting.
Look in Google Search Console for a specific signature: clicks and click-through rate falling while impressions and average position hold steady or rise. That means you still rank — Google still shows your page — but the click is being intercepted by the answer above you, which is different from a normal ranking loss where position gets worse. Filter to your informational, question-shaped queries (the ones AI Overviews trigger on most) and compare against a pre-2024 baseline. Search Console now also reports AI Overviews and AI Mode impressions specifically, so you can see when Google is surfacing you inside an AI answer.
Kompozy is an AI content generation and multi-platform publishing engine, not a traffic tracker. Its role fits the two things the data says actually help: being a cited winner rather than an uncited loser, which rewards consistent, corroborated presence across the surfaces AI answers pull from; and reducing dependence on the Google click entirely. From one Persona Brief it generates a citable blog article plus the same message as social posts, images, short-form video, and an email newsletter, then publishes across the eight social platforms plus blog and email behind a review gate — presence on the surfaces where a click to your site is not the required step.
Google says AI Overviews keep total click volume "relatively stable" and send "slightly more quality clicks" than before. Independent data disagrees: Pew found clicks fall to 8% from 15% when an AI summary appears, and Chartbeat measured publisher search traffic down about a third year-over-year. Both are partly true — AI Overviews concentrate the surviving clicks on cited brands while draining smaller publishers. Measure your own traffic by query intent in Search Console, not the headline numbers.
Get started → · ← All guides · Compare Kompozy vs other tools