What to know
- Keep a record of the analytical choices behind an AI-assisted result.
What the sessions show
Anthropic’s November 5 study analyzed 207,049 sessions classified as scientific research between August 2 and September 30, 2026. Chat and Cowork accounted for 128,914 sessions, with 78,135 from Claude Code. After weighting for sampling, data analysis made up 33.7% of Chat and Cowork research sessions and 48.1% of Claude Code research sessions.
The sample covers Free, Pro and Max plans and includes only interactive use of Claude Code. Language models identified scientific work and assigned tasks. The main analysis retained sessions with both a scientific field and a research task. The shares describe Claude usage in this sample. Scientific output, productivity and employment would require separate measures.
Sources: Anthropic — Research-session study; Anthropic — Methods appendix
Our interpretation
For a research portfolio, document an analytical decision alongside the code. Explain why a comparison is valid, which observations you excluded, or what evidence would change your conclusion. We suggest this exercise as a way to make your reasoning easy for a reviewer to inspect.
Sources & method
AI-generated synthesis of the study and appendix. The practice suggestion is editorial interpretation.
- Research-session study Anthropic ·
- Methods appendix Anthropic ·
Drafted and source-checked by Codex AI agents. Read our sourcing, illustration, and correction policy.
Explore the occupations
What could this mean for your work?
See task exposure and potential career moves for roles connected to this story.
