Google AI Overviews Study Finds Users Forget the Sources Cited

The click was never the whole story

For two years, the argument over Google's AI Overviews has mostly been about clicks. A new experiment from Forbes Advisor looks at a second loss that publishers may find harder to measure and harder to win back: whether people remember who supplied the information at all.

The study recruited 796 U.S. adults and split them into two groups. One group saw Google results pages with an AI Overview and the other saw the same pages without one.1 With traditional results, about 76% of participants could name at least one website they had seen. With an AI Overview, that fell to 56%, and the average number of sources people could recall dropped from 1.65 to 1.10.1 The researchers call this a "citation blind spot." They also found that community platforms such as Reddit and YouTube stayed more memorable than traditional editorial publishers.1

Intent to click fell too. With an Overview present, 67% of participants said they would open a link, compared with 84% without one. Another 19% said they would rely on the AI summary instead of visiting the primary source.1 Dr. Elena Hill-Artamonova, who led the study, put it bluntly: the Overview takes both the credit for the answer and the click.1

The study also tested Google's claim that searchers like the feature. It found no meaningful rise in trust, usefulness or satisfaction. Trust was already high, at about 93%, in both groups.1 Gen Z was the exception. Its trust rose from 49% to 67% when an Overview appeared, while its source recall fell from 74% to 60%.1

Why the memory finding matters more than another CTR number

The click damage is already well documented. What this study adds is evidence that being cited is losing value even for the brands that get cited.

This cuts against a common argument in the answer engine optimization (AEO) industry. Vendors and agencies have been telling brands that zero-click visibility still builds recognition and trust when nobody visits.11 HubSpot describes AEO's goal as raising brand awareness and consideration "even without a click."19 Similarweb's framework counts zero-click impressions as a core metric.13

The Forbes data suggests that premise is weaker than it sounds. If a fifth fewer users can name any site after seeing an Overview, a citation inside it delivers less brand exposure than a ranking once did. Editorial publishers seem to suffer more than platforms with built-in community identity.1 An SEO consultant quoted in the study, identified only as Shepard, said a page can now rank first, appear in an Overview and still get very few visitors. In his view, the visibility that counts now is being recommended or being searched for by name.1

The study's limits deserve weight. Participants saw static pages rather than live ones. The study measured intended clicks, not actual ones. The sample is not nationally representative, and the Boomer group had only about 21 or 22 people.1 The authors say the result needs to be replicated.1 Still, its direction matches better-controlled research, which is why it deserves attention.

The causal evidence has caught up with the complaints

The strongest support comes from a randomized field experiment by Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon. They built a Chrome extension that hid AI Overviews for a randomly chosen group of real users browsing normally.22 When an Overview appeared, outbound organic clicks fell by about 39.8% and zero-click searches rose. Clicks on ads and on Google's own pages did not change significantly.22 In other words, the Overviews replaced visits to publishers rather than moving attention elsewhere on the page.

The effect was concentrated where Overviews appeared at the top of the page, which was more than 87% of the time. In those cases, removing the Overview produced an 88% relative increase in outbound clicks.22 The study also found no measurable change in satisfaction or ease of finding information.6 On that point, the field experiment and the Forbes survey reach the same conclusion through different methods: users lose little by default when Overviews are removed, and publishers lose a lot when they are shown.

Other research points the same way. Pew's tracking of 68,879 real searches found that users clicked a traditional result in 8% of visits with an AI summary, versus 15% without one. They clicked links inside the summary in only 1% of visits.4 Users were also more likely to end their browsing session after a page with a summary: 26%, versus 16%.4

Where the numbers diverge

The research agrees on the direction but not on the size, and the gap is worth understanding.

Ahrefs first estimated a 34.5% drop in position-one click-through rate across 300,000 keywords.3 A follow-up comparing December 2023 with December 2025 put the figure at 58%. Pages in position two lost about half their clicks, and even position ten lost nearly 20%.8 Amsive's analysis of 700,000 keywords found a smaller average decline of 15.49%, with steeper losses in some cases.2 A difference-in-differences study of Wikipedia found that Overviews cut daily traffic to English articles by about 15%.29 Seer Interactive reported organic click-through rate falling from 1.76% to 0.61% on queries with Overviews.10

These studies measure different things. Some look at CTR at a single ranking position, some at traffic per page, some at clicks per search. Their samples range from encyclopedia pages to commercial keywords. Query type matters a great deal. In Agarwal and Sen's data, informational searches showed large effects, while navigational and transactional searches showed none that could be told apart from zero.22 Branded queries are the main exception: Amsive found click-through rates rose 18.68% when an Overview appeared on one.2

Publishers' own figures are worse than the averages. Chartbeat data cited by the Reuters Institute shows U.S. organic search traffic across 2,500 publisher sites fell 38% over a year. Small publishers lost 60% of search referrals and large ones 22%.23 A separate Chartbeat report found news sites lost 40% of Google traffic between July 2025 and July 2026.27 DMG Media told the UK's competition regulator that click-through rates fell by up to 89% in its worst cases.21

Google has consistently disputed this research. It called Pew's methodology flawed and its queries unrepresentative, and said it has not seen significant drops in overall web traffic.4 The company also rejected Authoritas's analysis as inaccurate.5 But Google has not published the data behind its own claims of higher-quality clicks, so its rebuttals are hard to test.21 With randomized studies now available, the burden of proof has shifted toward Google.

Google's response: more links, still no click data

The timing is pointed. On October 15, days after the Forbes study appeared, Google announced five changes to links in AI Mode and AI Overviews. Inline links will appear higher in answers and more often. Google is testing a top-of-answer section for the one or two official pages a task requires. News carousels are expanding to more markets, related-article links are getting thumbnails, and Google is testing clickable creator names on video results.3133 The company says use of its Preferred Sources feature has grown more than sevenfold this year, though it gave no baseline.33 More than 400 publications have joined its paid News AI licensing pilot, including Axios, CNN and the Financial Times.32

What Google left out is just as telling. It still does not report click data for AI features. Search Console gives every link inside an Overview the same position as the Overview itself, so site owners cannot see whether a higher inline link actually helps them.33 Search Engine Roundtable's Barry Schwartz noted that this is at least the latest of many link redesigns, none of which has shown a lasting effect on clicks.36 Google made a similar set of changes in May, while publishers were already filing antitrust complaints.24

These changes target clicks, not memory. A thumbnail may draw a tap. It does nothing about users failing to notice whose reporting supplied the answer.

What this means for SEO and AEO

The AEO industry is growing quickly. Adobe now sells a brand-visibility platform built on Semrush's database of AI prompts.17 Yet its own guidance does not agree with itself. One guide says 99% of AI Overview citations come from pages ranking in the organic top 10.15 Compiled benchmarks claim 83% come from outside the top 10, and that overlap between Overview citations and top-10 rankings fell to between 17% and 38% by early 2026.1214 Authoritas found that about 70% of pages cited in Overviews changed within two to three months.21 When the basic numbers conflict this much, AEO metrics should be treated as rough signals, not reliable measurements.

The practical takeaways follow from the evidence. Brand demand is the one area where AI Overviews add clicks instead of removing them.2 Informational content faces the most risk.22 Blocking Google's crawlers is not a solution either: one study found sites that block Google's AI crawler are less likely to be cited, even though the only way to opt out of Overviews is to leave Google Search entirely.29

The Forbes experiment adds that being cited does not guarantee being remembered. For publishers, the clicks are going, and recognition from the readers who stay inside Google is going with them. More inline links will not fix that. Fixing it would take independent measurement and payment terms that make Google's use of publisher content worth more to publishers than the attention it takes away.