Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia

Search engines increasingly display AI-generated answers above organic links, potentially displacing traffic to upstream publishers. We estimate the impact of Google’s AI Overviews (AIO) on Wikipedia’s search traffic using AIO’s staggered geographic rollout and Wikipedia’s multilingual structure. Our difference-in-differences design compares monthly external-search referrals to English Wikipedia articles with referrals to the same articles in German and French, and finds that default AIO availability reduced English search traffic by 5.45% and 4.82% , respectively. Our results suggest that answer-producing digital intermediaries can materially reallocate attention away from informational publishers, with implications for content monetization, search platform design, and policy.

Introduction. Digital information intermediaries, such as search engines and content aggregators, have traditionally created value by organizing third-party content, lowering search costs, and directing users to relevant publishers. Generative AI changes this role by allowing search engines and information intermediaries to synthesize information from multiple upstream publishers and present a self-contained response within their own interfaces. Rather than merely helping users locate relevant content, these intermediaries can now satisfy publishers whose content supports the generated response. Because many publishers rely heavily on search referrals, even modest reductions in referral traffic can affect advertising revenue, subscription acquisition, producing information intermediaries complement or substitute for upstream content producers is therefore Google’s AI Overviews (AIO) provides a prominent setting to study this question. Google’s AIO feature presents a synthesized answer prominently on the search-results page, before the traditional organic links, There is substantial disagreement over the effect of AIO on upstream publisher traffic. Publishers argue that by placing a synthesized answer above the organic results, AIO may satisfy users’ information needs without an external click and push source pages farther down the search-results page. In this view, AIO acts as a substitute for publisher content and cannibalizes traffic, even when it cites the underlying sources (Davies 2025; Miller 2025). Google counters that AIO is complementary: it lowers search costs, improves The economic and policy stakes of this disagreement are substantial. U.S. search advertising generated $102.9 billion in revenue in 2024, accounting for nearly 40% of all U.S. internet advertising revenue (Interactive Advertising Bureau (IAB) 2025). Google has also steadily integrated advertising into AIO,

Related work. At a high-level, this debate echoes prior work on news aggregators, which asks whether aggregators However, AIO changes the mechanism in a way that makes direct extrapolation difficult. Unlike an aggregator

Method. We estimate the causal effect of Google’s AI Overviews on publisher search traffic, using Wikipedia as a large, transparent, and economically important empirical setting. Wikipedia is particularly well suited to this analysis for three reasons. First, it is one of the world’s most visited websites and a major source of informational content, receiving more than 130 billion page views per year (Wikipedia 2025a,b). Search Wikipedia potentially easy to replace. On the other hand, Wikipedia ranks prominently in organic search and AIO rollout and analysis period. Google’s rollout of AI Overviews was staggered across countries and Treatment and control languages. The Wikimedia clickstream data report search traffic by language edition and article, but not by the reader’s country. We therefore proxy geographic exposure to AIO using Wikipedia language editions. We designate the English-language edition (en.wikipedia.org) as the treated group because approximately 40% of its traffic originates in the United States, where AIO became part of the default search experience during our treatment period. For the same underlying articles, we use 3.1 Model-Free Analysis

Discussion. and Implications Scope of the estimates. We next clarify what our estimates capture. Our outcome measures visits that arrive directly from external search engines. We therefore interpret the estimates as the first-order effect of Publisher monetization and content supply. Wikipedia does not monetize its traffic through advertising or subscriptions, but many publishers rely directly on search-originated visits for advertising impressions, subscription acquisition, lead generation, and commerce. For these publishers, a decline in search traffic Intermediary design and welfare. Our findings highlight a fundamental difference between link-based Platform and policy implications. Our findings also raise broader questions about market power, attribu-

Conclusion. Generative AI is changing search engines from intermediaries that organize links into systems that answer users’ questions within the search interface. We study the publisher-side consequences of this shift by combining Google’s staggered rollout of AI Overviews with Wikimedia’s article-level clickstream data.

Our findings imply that generative-answer features such as AIO in search engines can materially reallocate attention away from publisher pages, even when it cites and links to that content. Although this shift may lower users’ search costs and improve access to information, it can also reduce publishers’ monetization opportunities and weaken the incentives that sustain the underlying content ecosystem. As digital intermediaries increasingly produce answers rather than merely organize links, preserving the incentives to create