GEO Times
A record of who generative AI search recommends
Home / International
International

95% of ChatGPT Users Use Google, Queries and Clicks Fell

Similarweb, Bocconi University, Pew and arXiv data show user counts, queries and clicks moving apart.

Source Search Engine Journal · 2026-09-04 · Checked 2026-09-05

Search Engine Journal gathered several studies published this year and concluded that user counts, query counts and click counts move in different directions. The phenomenon of steady Google impressions alongside falling traffic comes from the gap between these three metrics.

In Similarweb's tally, the share of ChatGPT users who are also Google users was 95% in September 2025 and stayed at 95% through May 2026. Visits to generative AI platforms rose 70% year over year over the same period. Similarweb read this as AI being added on top of search rather than replacing it.

However, this overlap metric counts anyone who searched Google even once in a month as a Google user. It excludes API access, desktop apps and AI features built into other apps. In a Pew survey of 5,119 U.S. adults in February, 49% said they had used a chatbot, 42% said they use one to find information, and 60% said they read the AI summary at the top of search results.

Queries Fell, Clicks Fell More

A Bocconi University working paper used Comscore's U.S. desktop clickstream data from October 2024 through July 2025. Traditional search queries fell by 3.14 per week among households that gained access to ChatGPT Search, about 9.4% of the prior average of 33.51 queries. After 20 weeks, the decline widened to 17.0%.

A comparison of logged-in versus anonymous users showed smaller declines, 4.9% and 8.2% after 20 weeks. Search traffic fell most for academic sites at 32.8%, reference sites at 26.5%, developer sites at 15.1% and news sites at 13.4%, while marketplace and entertainment sites barely moved. Google's household reach in the same panel held steady at about 49% while ChatGPT's rose from 4.6% to 7.9%, though the authors stress these are descriptive figures, not causal ones.

A working paper posted to SSRN in February by Gholami and colleagues covered 4,394 Comscore desktop users who started using an LLM between August 2023 and January 2024. After adoption, the number of sites visited weekly rose, while traditional search rose temporarily before later declining. The authors said their search estimates carry substantial noise.

Numbers Google presented at I/O in May point the other way. AI Overviews has more than 2.5 billion monthly active users and AI Mode passed 1 billion in its first year, and Sundar Pichai said "people who use the AI features in Search use Search more." Search Engine Journal noted there is no baseline data available to independently verify that claim.

A field experiment posted to arXiv on August 18 by Wang, Gleason, Bart, Wilson and Metaxa observed 1,100 U.S. Chrome users whose main search engine is Google for three days, then split them into three groups for a week: regular Google, Google with AI Overviews hidden, and a group whose searches were all routed to AI Mode. AI Mode usage, which was 0.6% at baseline, reached 94.7% in the last group.

Under this condition, the click-through rate to external sites fell by 18.8 percentage points. The 95% confidence interval ran from -22.2 to -15.3. Daily search sessions fell by about 0.92 from an average of 4.0, and session time rose by 0.43 minutes.

The share of participants who clicked to news sites fell by 12.5 percentage points, Reddit by 21.2 points and Wikipedia by 9.9 points. The share using Bing, DuckDuckGo or Yahoo rose by 11.2 points, even as ratings of trust, satisfaction and usefulness declined. Among 309 participants who answered an open-ended question, 47, about 15.3%, said it was hard to reach the site they wanted.

In the group with AI Overviews hidden, external clicks rose by 8.8 percentage points among those for whom the extension actually worked. The authors also noted that their sample skews younger and more educated than the U.S. population overall, and that seven days is a short window for observing habit change.

In Korea

All of the research above is based on U.S. user data. No public study of Korean-language users at a comparable scale appears in this article. Korean site operators who want to check for the same pattern can compare their Search Console impression and click trends side by side.

Source: Search Engine Journal, Matt G. Southern, September 4, 2026. Checked September 5, 2026. Read the original