How Generative AI Disrupts Search: An Empirical Study of Google Search, Gemini, and AI Overviews
Researchers benchmarked how generative AI search fundamentally alters information retrieval compared to traditional engines. Using 11,500 real queries, they found AI Overviews appear above organic results in over half of cases, with controversial topics triggering AI-generated summaries at higher rates. This empirical study reveals a structural shift in how users encounter information: AI systems now mediate and reframe search results before users see traditional links. The findings matter for understanding whether generative search improves discovery or concentrates authority in LLM-generated abstracts, with implications for publisher traffic, search equity, and how misinformation spreads through AI-filtered interfaces.
Modelwire context
Analyst takeThe study's most underreported finding is that controversial topics trigger AI Overviews at disproportionately higher rates, meaning the topics where source diversity and attribution matter most are precisely where a single LLM-generated summary is most likely to displace the underlying links.
The misinformation angle here connects directly to the MM-StanceDet paper covered the same day, which tackled how AI systems reconcile conflicting signals across modalities. That work was framed as a solution to content moderation problems, but this search study suggests the upstream problem is structural: AI Overviews are already mediating contested information before any stance-detection layer gets involved. Meanwhile, the broader question of whether AI systems concentrate epistemic authority is one this site has been circling, from LLM-assisted peer review to AI-measured research impact. The pattern is consistent: AI is inserting itself as an intermediary at every point where humans previously exercised direct judgment over sources.
Watch whether Google's publisher traffic data, if disclosed in any upcoming earnings call or antitrust proceeding, shows a measurable click-through decline correlated with AI Overview frequency. That would convert this empirical study from a structural observation into a documented economic harm with regulatory weight.
Coverage we drew on
This analysis is generated by Modelwire’s editorial layer from our archive and the summary above. It is not a substitute for the original reporting. How we write it.
MentionsGoogle Search · Google Gemini · AI Overviews · Gemini Flash 2.5
Modelwire Editorial
This synthesis and analysis was prepared by the Modelwire editorial team. We use advanced language models to read, ground, and connect the day’s most significant AI developments, providing original strategic context that helps practitioners and leaders stay ahead of the frontier.
Modelwire summarizes, we don’t republish. The full content lives on arxiv.org. If you’re a publisher and want a different summarization policy for your work, see our takedown page.