Google's AI-in-science study mines 15M Gemini interactions, 2,600 models and 600-scientist survey

danielrock · x · 2026-09-15

Google published an empirical study on how scientists actually use AI, led by mcodreanu and promoted by Daniel Rock. It draws on three data sources: 15 million Gemini interactions, 2,600 specialized AI models, and a survey of 600 scientists.

The work maps where AI enters the scientific workflow and how adoption varies across fields—one of the largest AI-for-Science studies to date, with more results promised. Details are in Arthur Turrell's thread and the original report.

Related event: Google Paper: 15M Gemini Interactions Show AI's Impact on Science(3 posts)→

Original post →

More from Research

Research channel →