Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia
New research provides the first causal evidence that AI search summaries directly reduce publisher traffic.
A landmark study provides the first causal evidence that AI-generated search summaries significantly reduce traffic to information publishers. Researchers Mehrzad Khosravi and Hema Yoganarasimhan analyzed the impact of Google's AI Overviews on Wikipedia traffic using a sophisticated difference-in-differences design. They leveraged the feature's staggered geographic rollout and Wikipedia's multilingual structure, comparing English articles exposed to AI Overviews with the same articles in Hindi, Indonesian, Japanese, and Portuguese editions that weren't exposed during the observation period.
The analysis of 161,382 matched article-language pairs revealed that AI Overview exposure reduces daily traffic to English Wikipedia articles by approximately 15%. The effects showed significant heterogeneity: Culture articles experienced the largest relative declines, while STEM articles saw substantially smaller impacts. This pattern suggests stronger substitution occurs when short synthesized answers can satisfy informational intent, particularly for topics where comprehensive source reading might be less essential.
The findings have substantial implications for the evolving search ecosystem. While platforms like Google argue AI summaries complement source pages by directing users through included links, this research indicates they often substitute for them instead. The 15% traffic reduction represents a material reallocation of attention away from informational publishers, raising questions about content monetization, search platform design, and potential policy interventions. As AI-generated answers become more prevalent in search interfaces, publishers may need to adapt their strategies for visibility and revenue in an answer-first search paradigm.
- Google AI Overviews caused a 15% average reduction in daily traffic to English Wikipedia articles, based on analysis of 161,382 article-language pairs.
- Traffic impact varied by topic: Culture articles saw the largest declines while STEM articles experienced substantially smaller effects.
- The study used a rigorous difference-in-differences design comparing exposed English articles to unexposed versions in four other languages during the staggered rollout.
Why It Matters
This provides concrete evidence that AI search summaries directly reduce publisher traffic, impacting content monetization and search ecosystem dynamics.