Anthropic measured how scientists actually use Claude

Summary

Anthropic published an analysis of 207,049 scientific Claude and Claude Code sessions from August and September 2026, using privacy-preserving classification. Formal sciences such as computer science, mathematics, and statistics showed higher use relative to researcher counts, while lab-based and medical fields showed less. Usage concentrated on data analysis and writing rather than on idea generation or data gathering, the tasks scientists themselves rank as most important. The authors estimate that wide adoption of current capabilities could double research output in some fields, and that a scenario in which AI handles all non-physical research work could lift computer science and statistics up to 15-fold and bench sciences three- to fourfold. The scenario figures are the company's own model-based estimates.

Why it matters for our work

The measurement shows AI adoption accelerating exactly where models are strongest while leaving gaps elsewhere. About 40% of surveyed researchers who actively use AI expect AI to carry out full research projects by 2030, a useful signal for planning research teams and workloads.

Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.