AI NewsGo-to-marketReported

AI Overviews cut outside clicks 40 percent in randomised test

A RevScope review finds AI Overviews cut clicks to outside sites by 40 percent in a randomised test.

AI News

Editorial3 min read

LinkedInX
Reveneau hero illustration of search traffic and AI answers

Why it mattersA team paying for growth through Google clicks is losing a large share of those clicks. No GEO technique has been shown to make up the gap.

The team at RevScope has pulled the recent measured studies of AI Overviews into one post. It went up on 9 November. Hacker News has it at 353 points. The headline is a 40 percent drop in clicks to outside sites. A randomised experiment produced that figure. For a team funding growth from Google clicks, that is the central reading.

The experiment is by Kartik Agarwal and Rachit Sen, posted to SSRN in June. The sample is 1,065 participants, with 770 placed in comparison groups. On a search with an Overview, participants made 0.37 clicks to outside sites on average. Without one, they made 0.62. That is a 39.8 percent reduction. Searches ending with no outbound click rose from 54 percent to 73 percent. Organic traffic in the experiment fell 18.5 percent overall.

How often Overviews appear

RevScope links the latest DataForSEO Search Insider, dated October 2026. DataForSEO tracks 190 million keywords. In September 2024, 5.93 percent of US keyword results showed an AI Overview. In August 2026 that figure was 61.43 percent. Informational queries carry one 70 percent of the time. Commercial queries carry one 51 percent of the time. Featured snippets, the old answer box, fell from 8.96 percent to 0.44 percent over the same window. RevScope notes DataForSEO sells search data and the figure is the vendor's own.

What happens inside the result page

Desktop click-through for the first organic result fell from 17.83 percent to 7.43 percent in one year. The measurement is RevScope's own, against DataForSEO's dataset. Mobile at position one fell from 25.17 to 20.80 percent. Pew Research ran a 2025 study of 900 US adults. On a page with an AI summary, 8 percent of people clicked a result. Without a summary, 15 percent clicked. People also stopped browsing more often, 26 percent against 16 percent.

What citations correlate with

SISTRIX studied 48,527 cited pages paired against uncited ones. A current year in the title raised citation likelihood by 2.7 times. A jump-link table of contents raised it 1.7 times. FAQ structured data raised it 1.5 times. The Aggarwal GEO paper from KDD 2024 ran 10,000 queries. Adding quotations lifted visibility about 41 percent. Adding statistics lifted it about 31 percent. Citing sources raised it about 27 percent. Keyword stuffing lowered it by 8 percent. SE Ranking looked at 300,000 domains and found no correlation between llms.txt files and citations.

The honest limit

RevScope is open about the gap in the data. Nearly all the large datasets come from vendors who sell search tools. The experiments are small and most of the observational data is correlational. AI assistant traffic across 74,752 sites in Ahrefs' March 2026 study was 0.28 percent of the total. Google is sending fewer clicks than before, and no AI tool is sending enough to replace them.

The practical move from this synthesis is the one RevScope ends on. Measure each channel by the revenue it produces. Rank and visit count are now the wrong measure. Fewer clicks are coming from search, so channel quality matters more than channel volume. GEO, as it reads today, looks like SEO plus two things: being present on the sites that get cited, and writing content a model can quote.

Source

Google's AI Overviews cut clicks to websites by 40 percent, RevScope. The Agarwal and Sen paper is at SSRN abstract 6513059.

Reported byRevScope

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project