Scientists using Claude turn to it most often for data analysis and writing, according to an Anthropic study of 207,049 research sessions. Use was especially concentrated in fields such as computer science, mathematics and statistics, where much of the work already happens in code or text.
What researchers bring to the chatbot
Anthropic studied sessions on consumer plans between August 2 and September 30, 2026. The main sample included 128,914 Chat and Cowork sessions and 78,135 Claude Code sessions. Data gathering and generating hypotheses accounted for relatively little use compared with the time researchers spend on those tasks.
A survey of 528 scientists broadly echoed the pattern. That group was made up largely of researchers in life and health sciences who already actively used AI, so it offers a view of early adopters. The session count also measures interactions, and a single researcher could account for several sessions.
What universities can learn from the study
For research teams weighing whether to pay for AI tools, the results point to specific jobs they can evaluate, such as making a chart or drafting a report. Anthropic measured usage, so questions about time saved, accuracy and the quality of the resulting research still need separate evidence.
