Google AI Overviews Click Losses Deepen as AI Mode Expands

The click gap is now well documented

For about two years, the argument over Google's AI Overviews was a fight between publisher anecdotes and Google's reassurances. In 2026 that changed. Randomized experiments, large clickstream panels and multi-billion-impression datasets now point the same way. When Google puts an AI-generated summary above the results, fewer people click through to the open web. Studies disagree on how many fewer. They do not disagree on the direction.

The most cited figure comes from Ahrefs. Its analysis of 300,000 keywords found that top-ranking pages on queries with an AI Overview get a 58% lower click-through rate, up from a 34.5% reduction measured eight months earlier25. Put simply, of every 100 clicks that once went to the top result, Google now keeps 5825. In absolute terms, position-one CTR for those keywords fell from 7.3% in December 2023 to 1.6% in December 202529. In Germany, SISTRIX found a similar pattern: top-result CTR fell from 27% to 11% when an overview appeared, which it estimates costs about 265 million organic clicks a month across the market2422.

Correlational studies like these are open to the objection that something else could explain the drop. The stronger 2026 evidence comes from experiments. In a randomized field study of 1,065 desktop Chrome users, hiding AI Overviews raised outbound organic clicks from 0.37 to 0.62 per search, and cut the chance of a search ending with no click from 0.73 to 0.543. The authors attribute a 39.8% fall in clicks to the feature, and found that sponsored clicks did not move3. The effect sat almost entirely in the top-of-page slot, where more than 87% of overviews appear3. A second preregistered experiment with about 1,100 US users found that hiding overviews raised click-through by 8.8 percentage points24. Neither study found any detectable gain in user satisfaction or trust when overviews were shown243.

This undercuts Google's main defense. The company has said overall organic clicks are "relatively stable" and that the clicks it sends are higher quality24. Industry observers note that Google has never published metrics to back up the quality claim1. The experiments do lend Google some support on one point: the June revision of one paper found no significant differences in bounce rate or time on site24. That means the clicks that remain are not worse. There are simply far fewer of them.

Where the numbers diverge

The biggest disagreement in the coverage is about how often AI Overviews appear at all. Semrush's 10-million-keyword panel put the share at 15.69% in November 2025, down from a July peak of 24.61%24. Conductor's enterprise panel measured about 25%24. BrightEdge put the figure near 48% in commercial verticals, with healthcare at 88%24. A Similarweb clickstream figure reported in July 2026 claimed 43% of US searches, though no methodology was published24. An academic benchmark found overviews on 51.5% to 65.6% of representative queries24.

These numbers look contradictory but mostly measure different things. Trending and news queries rarely trigger overviews. NewzDash found them on just 3.47% of US trending news searches24. Question-style and comparison queries trigger them most of the time28. Any single "share of searches" figure depends heavily on which queries are in the sample.

The second disagreement is about whether the damage is getting worse or levelling off. Seer Interactive has the longest-running dataset: 5.47 million queries and 2.43 billion impressions. It found organic CTR on AI Overview queries fell from 1.76% in mid-2024 to 0.61% in September 2025, then rose to 2.4% by February 202621. Some commentators read this as a recovery. A more careful reading is that the gap is narrowing but not closing. CTR on queries without an overview climbed to 3.8% over the same period, so a gap of roughly 37% remains21. One plausible explanation is that overviews now absorb the low-intent queries that were never going to produce clicks29. SparkToro's figures point the other way: zero-click searches across all US Google searches rose to 68% in early 2026, from 60% in 2024. On balance, the evidence supports a permanently lower baseline rather than a rebound.

Publishers are taking the hit

For publishers, these percentages show up as layoffs. Chartbeat data shows search referral traffic falling 60% for small publishers, 47% for mid-sized ones and 22% for large ones over two years. Referrals from chatbots grew more than 200% but still account for less than 1% of page views24. Piano's benchmarks across hundreds of sites show a 36% drop in search traffic8. Growtika tracked ten major tech outlets whose combined Google traffic fell from 112 million monthly US visits to under 50 million. Digital Trends alone dropped 97%10. Future has reportedly cut jobs a second time in 2026, with Tom's Guide's UK audience down 55% since before AI Overviews launched24.

The damage is uneven. Chartbeat's figures show smaller sites losing ground nearly three times faster than large ones9. Amsive found that branded queries actually gained CTR when an overview appeared, while non-branded queries lost nearly 20%24. Health, tech and service content (weather, horoscopes, how-to pages) are the most exposed. Breaking news and sport are largely protected248. The likely result is a two-tier web: established brands with direct audiences hold up, while independent informational publishers lose the referral traffic their business models depended on1.

The legal pressure has grown with the traffic losses. Penske Media is pursuing an antitrust case accusing Google of "cannibalizing" publisher traffic. The European Publishers Council has filed a complaint with the European Commission. Chegg has sued over what it says was a 49% collapse in non-subscriber traffic245.

Google's response: link tweaks and a merger with AI Mode

Google's product changes this year pull in two directions. On May 6, it announced five changes: inline links next to individual claims, hover previews on desktop, "Subscribed" labels for publications a user pays for, article suggestions at the end of answers, and a section of forum and community quotes1516. On October 15 it went further, moving inline links higher in responses, testing a section for official sites on navigational queries, and expanding article carousels worldwide20. In June, Google gave site owners a Search Console switch to opt out of AI features and added reporting on AI impressions13.

The structural changes are heading the other way. At I/O in May, Google combined AI Overviews and AI Mode into one flow running on Gemini 3.5 Flash. It said AI Mode had passed 1 billion monthly users and that AI Overviews reached more than 2.5 billion13. Since August 28, some overviews expand automatically into full AI Mode answers, pushing organic results further down the page12. Since late September, some links inside overviews open an AI Mode conversation instead of a website12. Search Engine Roundtable's Barry Schwartz said bluntly that the change is bad for publishers13. Early French data after the July launch showed sites with heavy overview exposure losing 23.1% of CTR within days12.

Our read: the link improvements make citations more visible, but they do not change the basic problem. A complete answer makes the click optional. Pew found that only about 1% of users click links inside an AI summary15. Making those links more prominent helps at the margins. Turning overviews into entry points for AI Mode works against that, at a much larger scale. One caveat applies: SparkToro's panel found AI Mode accounted for only 0.34% of US searches in early 2026, despite the billion-user figure13. The merger may matter more for where Google is heading than for traffic today.

What answer engine optimization means now

As organic clicks shrink, the industry has moved toward "answer engine optimization" and "generative engine optimization." Google's own guidance rejects the distinction. Its new documentation says optimizing for generative AI search "is still SEO," and says llms.txt files, content chunking and special schema are unnecessary14.

The data shows why the new labels caught on anyway. Ranking and being cited have separated. One analysis found top-10 results made up 76% of AI Overview citations in mid-2025 but only about 38% by early 202611. Another found nearly 30% of cited domains do not appear anywhere on the first page for the same query24. Citation is valuable. Per million impressions, cited pages earn about 20,743 clicks, against 9,445 for uncited pages21. But cited pages still fall short of the 33,500 clicks a query earned with no overview at all21. A citation reduces the loss. It does not make it up.

The practical conclusion is clear. Optimizing for citations is now a basic requirement for publishers, but it cannot replace direct relationships with readers. The publishers still growing revenue are the ones building apps, newsletters and registration8. Google says it wants AI answers to send more clicks20. Until it publishes data showing that happening, publishers should plan for permanently lower search traffic.