Google let AI Overviews eat the whole page. Your organic click through rate just found out

What Google actually changed, and why it matters more than a redesign#

A fox reads the shape of a clearing before it steps into it. Search marketers now need to read the shape of a results page the same way, because Google has quietly changed what that shape can be.

For some queries, Google's own systems now decide an AI Overview should dynamically expand to fill the entire results page rather than sitting in a fixed box above the ten blue links, a statement Google gave the press and Search Engine Land reported around the end of August 2026. Google's own framing is that this happens "for some queries, AI Overviews may dynamically expand for topics where our systems determine it's most useful for people."

Read that sentence the way a fox reads a scent trail: the decision is automated, the trigger is undisclosed, and the effect is that organic click through rate on an affected query can drop to something close to zero before a single blue link is visible on the first screen. This is not a ranking change. Nothing about where your page sits in the index moves. What moves is whether a human ever scrolls far enough to see it.

Why this is not just another AI Overviews update#

Google has expanded AI Overviews coverage steadily since 2024. What is different about the full-page behaviour is that it removes the one thing that made a boxed AI Overview survivable for organic search: the ten blue links stayed visible on the same screen, competing for the same glance. An ai overview full screen expansion pushes them below the fold entirely on the queries it targets, which is a materially different competitive position than losing a few pixels of prominence.

Practitioners noticed the change within hours of it landing, and the framing that spread fastest through search marketing circles was blunt: ranking first under an ai overview full screen takeover reads to a searcher exactly like ranking on page two used to. The position in the index has not moved. The distance a human eye has to travel to reach it has, and that distance is now measured in scroll depth rather than blue-link rank.

Honcho's analysis of the wider pattern puts it plainly: impressions rise as AI Overviews spread, because the query still counts as a search, while clicks fall on exactly the queries where the summary now answers the question outright. That divergence, not any single algorithm update, is the actual story behind a falling organic click through rate this year.

What the organic click through rate studies actually found#

Most of what circulates about AI Overviews and clicks is observational: traffic before a rollout compared with traffic after, with every other variable in the market moving at the same time. That kind of evidence is suggestive, never conclusive, because a hundred other things also changed between the two dates.

The study that changed the argument is different in method, not just in headline number. Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon University's Heinz College built a browser extension that randomly assigned real US-based participants to see or not see AI Overviews on identical searches, then measured what each group actually clicked, per the paper posted to SSRN. Randomising exposure, rather than comparing before and after, is what lets the result claim causation instead of correlation.

The headline finding, covered independently by ppc.land and Search Engine Journal, is that when an AI Overview appears, outbound organic clicks fall by roughly 38 to 40%, and the share of searches ending with zero clicks anywhere rises from 54% to 72%. Two independent newsrooms reaching the same number from the same underlying paper is exactly the kind of corroboration organic click through rate claims usually lack.

The scatter above compresses each study to a single point, but the methods behind those points differ enough to matter. Agarwal and Sen randomly assigned 1,065 US participants to see or not see an AI Overview. Amsive Digital tracked roughly 700,000 keywords observationally and found a smaller 15.5% drop, closer to the low end of the range. Seer Interactive's longitudinal tracking, covered in a writeup by Karooya, found a steeper 61% drop over a longer window.

The spread between 15.5% and 61% is itself informative about organic click through rate: it tells you the effect depends heavily on query type, and averaging across all queries hides exactly the queries that matter most to a given site. A commercial query under an ai overview full screen takeover behaves nothing like an informational one with a small boxed summary, and treating them as one number is how a marketer talks themself out of noticing which of their own pages is actually exposed. Segment your own organic click through rate by query type before drawing any conclusion from an industry-wide average.

Impressions hold. Clicks do not.#

Google Search Console's own reporting mechanics make the shape of the problem visible if you know where to look. AI Overview impressions count as ordinary impressions in the Performance report, and clicks from inside an AI Overview are folded into ordinary organic clicks rather than broken out separately by default. A query where impressions hold steady while clicks fall is the signature of exactly this problem, not a ranking demotion, and it is the single fastest diagnostic available before you touch a page.

Five fixes that improve ai search optimization, not just SEO#

None of the five moves below require guessing at Google's undisclosed trigger for full-page expansion. They improve the odds of being the source cited inside the summary, which is the only form of visibility left on a query the summary has claimed entirely. Citation slots behave differently from ranking slots on the evidence gathered so far, and a page can rank ninth while still being the one quoted inside the summary itself.

This matters because organic click through rate is no longer a single number worth optimising in isolation. It is the visible half of a two-part metric, the other half being citation frequency, and a team that only watches the half that is shrinking will conclude the whole channel is failing when in fact the channel has simply split in two.

Read that shape the way a fox reads a hedgerow at dusk: the traffic has not left the field, it has changed which path it takes across it. Chasing the old path harder does not bring the click back. Building a page a generative system can safely quote does.

That is the whole shift in one sentence from a working SEO and GEO practitioner: ranking first no longer guarantees a click, but being the cited source still does. Building for citation is the practical core of seo for ai search right now, and it is a discipline built on top of classic SEO fundamentals, not a replacement for them. Crawlability, page speed and internal linking still decide whether a page is eligible to be summarised at all.

How to know if you have actually learned how to rank in AI Overviews#

Classic reporting assumes every impression that matters ends in a click. Full-page AI Overviews break that assumption on the queries they touch, so the measurement has to move upstream, from sessions to visibility and citation.

Set a baseline before touching a single page. Take a snapshot of impressions, clicks and average position for your twenty most commercially significant queries, note which already trigger an AI Overview, and repeat the sweep monthly. The pressure is not unique to search either: Cloudflare's own content-access policy, which by default restricts AI crawlers on many sites from 15 September 2026, is a parallel squeeze on the same content economics from the crawling side rather than the results-page side.

Pick one headline metric and defend it to whoever holds the budget: the share of your priority queries where you appear inside the summary, not merely ranked beneath it. It survives an algorithm change, and it turns a vague anxiety about organic click through rate into a line a board can actually watch move.

If you want that reporting layer built rather than described, that is what folkfox's SEO and GEO practice does, and it is the same discipline applied across content marketing and every regulated vertical folkfox works in, from fintech to healthcare.

Frequently asked questions#