AI for scientific discovery: How researchers use Claude

Researchers analyzed 207,000 Claude AI sessions to understand how scientists use AI for research, finding that Claude is most used for data analysis and writing tasks rather than idea generation—which scientists rank as most important for discovery. The study shows that Claude usage correlates with AI capability levels across different scientific fields, with formal sciences like computer science using it more than lab-based sciences. If AI capabilities are widely adopted and expanded to handle all non-physical research work, it could potentially double research output in some fields or increase it up to 15-fold in computational sciences.