Two-Thirds of Google Searches Now End Without a Click. Whether Your Website Gets Cited Matters More Than Whether It Ranks

A search on Google now ends without anyone clicking anywhere two times out of three. That is the headline from SparkToro's newest zero-click search study, built on Similarweb's US desktop and mobile clickstream panel for January to April 2026, which put the zero-click rate at 68.01%, up from 60.45% just two years earlier. The company is candid about the limits of the data, it covers the US only, excludes Google's own mobile app, and drew on a different clickstream provider than its earlier reports, so the exact percentage-point jump shouldn't be read as gospel. The direction, though, isn't in doubt, and nothing about UK search behaviour suggests it's moving the other way.

AI Overviews are doing most of the damage, though not all of it. SparkToro's analysis found the summaries now appear on roughly one in five searches, and when they do, click-through rates fall by close to 60%. Ahrefs data cited in the same report showed referral traffic share to tracked domains falling eight percentage points between June 2025 and May 2026. A search engine that used to be a doorway to somebody else's website is increasingly answering the question itself and keeping the visitor.

The more useful number sits one level down, inside the AI Overviews themselves, and it changes the practical question a business should be asking. Seer Interactive's 2026 update, tracking 53 brands across 5.47 million queries and 2.43 billion organic impressions between January 2025 and February 2026, found organic click-through averaged 3.35% on searches with no AI Overview at all, 2.07% when an Overview appeared and cited the site, and just 0.94% when an Overview appeared without citing it. Being summarised without credit costs more than half the traffic a normal search result would have earned. Being cited inside the summary keeps most of it.

There's a recovery worth noting too, and it cuts against the more alarmist version of this story. Seer's data shows organic click-through on Overview-present queries climbing from a floor of 1.3% in December 2025 to 2.4% by February 2026, an 85% jump in ten weeks. Google appears to be tuning how much of an answer the Overview gives before nudging people onward, and that tuning moves. Paid click-through, for what it's worth, barely moved across the same period, holding between roughly 14% and 18% regardless of what organic was doing. None of this supports either "SEO is dead" or "nothing has changed"; both ignore the actual shape of the data.

For a UK engineering or manufacturing SME, the practical shift is which question decides whether a page earns a click at all. Ranking well used to be most of the job. Now a page can rank respectably and still lose almost all its traffic if the Overview answers the query without naming the source, and the difference between being cited and being ignored comes down to whether the page states something specific enough for a model to lift and attribute. We wrote last month about how AI is changing the way buyers vet UK engineering and manufacturing suppliers, and the same logic runs the other way too: a page built around a genuine claim, a named project, a real figure, a plainly stated capability, is one both a buyer and a language model can actually use. A generic capability statement gives neither one anything worth quoting.

None of this is an argument for chasing every algorithm update instead of doing the underlying work. It's an argument for treating a page's citability, whether a model reading it can find a fact clean enough to quote and attribute, as seriously as its ranking position, because ranking well while staying invisible inside an AI answer is now a genuinely common outcome. A specification page that states a tolerance, a lead time or a certification standard in plain text a model can lift beats one that describes the same capability only in marketing language a model has to interpret first.

This is roughly the discipline Nebula applies when reviewing a client's website: work through which pages carry a specific, checkable claim and which are still describing capability in the abstract, then rewrite the second group first, because those are the pages losing the most ground to an Overview that has nothing to quote from them. Getting this right isn't really about chasing an algorithm. It's about making a website answer the same question a buyer, human or otherwise, is actually asking it.