A study published July 8, 2026 by researchers at Bocconi University found that expanding access to ChatGPT Search reduced traditional search engine queries by 9.4% on average, with the decline deepening to 17.0% after twenty weeks of exposure, based on desktop clickstream data covering more than 45,000 United States households. The paper, titled "Answering Without Referring: How AI Search Rewrites the Web's Economic Bargain," is by authors Qiaoni Shi, Kai Zhu, and Kai Gu, posted to the arXiv preprint server under the Computers and Society category. The research draws on Comscore desktop clickstream records spanning October 2024 through July 2025. Using an "information-seeking occasion" as the denominator, the study found that ChatGPT produced a clean outbound referral in only 5.2% of conversation sessions, compared with 31.1% of Google queries. Among 56,578 households that used ChatGPT, 74.4% never generated a single clean referral to an outside website, versus 9.6% of 291,747 Google-querying households. Restricting analysis to household-weeks where the same household used both platforms, with fixed effects, moved the gap from a raw 31.5 percentage-point difference to 29.0 points.