AI for scientific discovery: How researchers use Claude

Anthropic researchers analyzed 207,000 Claude and Claude Code sessions to map how scientists use its AI models, finding adoption concentrated in formal sciences like computer science, mathematics, and statistics, with less use in health and lab-based physical sciences. Data analysis accounts for roughly one-third of sessions on Claude and nearly half on Claude Code, while idea generation and data gathering—tasks researchers rank as most important for discovery—remain underused. The study, led by Johannes Hermle, Ben Jones, and Lukas Althoff, found that tasks AI performs best receive two to four times more use than researcher time allocation would predict, while physical tasks see little uptake. In a simple model, widespread adoption of current capabilities could double research output in some fields, and AI handling all non-physical work could boost output up to 15-fold in computer science and statistics and three- to fourfold in bench sciences like chemistry and biology. About 40% of surveyed researchers who actively use AI expect models to run full research projects autonomously by 2030. The findings matter as AI labs and funders weigh where to target scientific tooling; watch for specialized products like Claude Science, where idea generation already draws twice the share of use seen on Claude, to close the gap between what scientists use AI for and what they consider most valuable.