Study analyzes 207,000 Claude research sessions
Researchers used privacy-preserving methods to analyze 207,000 Claude and Claude Code sessions across scientific fields and research tasks. Relative to researcher counts, use was higher in formal sciences such as computer science, mathematics and statistics, and lower in health and lab-based physical sciences. Scientists primarily used the tools for data analysis and writing. The report found that Claude use tracked task-level AI capability, with the most AI-suitable tasks receiving two to four times more use than researchers’ time spent on them would predict. A simple model also explored scenarios in which broad adoption of existing AI capabilities could double research output in some fields; these estimates depend on the model’s assumptions.
The analysis provides a large-scale view of how scientists use Claude and Claude Code across disciplines and tasks, helping characterize the current state of AI use in research.
Johannes Hermle*, Ben Jones*, Lukas Althoff*, Ryan Heller, Thomas Lyttelton, Carolyn Stein, Nathan Wilmers, Maxim Massenkoff, Peter McCrory
*Lead authors of the report