# Anthropic Study Finds Data Analysis Leads Claude Research Use

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/50519
Published: 2026-11-07T02:36:54.000Z
Updated: 2026-11-07T02:36:54.000Z
Section: Technology

> Data analysis made up about one-third of Claude sessions. Claude Science entered beta on Oct. 14, after the study period ended.

Anthropic’s analysis of 207,049 research sessions found that scientists used Claude most often for data analysis, with adoption levels varying across disciplines. Relative to the number of researchers in each field, use was higher in computer science, mathematics and statistics than in health sciences and lab-based physical sciences.

The sessions were recorded between Aug. 2 and Sept. 30. Data analysis made up about one-third of Claude sessions and close to half of Claude Code sessions. Writing was also a frequent task. Researchers used Claude less often for generating ideas or gathering data, even though 528 survey respondents considered both activities important to scientific discovery. That split shows a difference between what scientists said mattered to discovery and the tasks most visible in Claude sessions.

Claude Science entered beta Oct. 14, after the session window ended. In that platform, a larger share of sessions focused on idea generation. It includes tools for displaying protein and chemical structures, producing outputs linked to code and working with high-performance computing environments.

Anthropic’s analysis also compared how often researchers used AI for tasks with estimates of how well AI performed those tasks. Tasks where AI performed well drew between two and four times as much use as tasks where it was less effective. The analysis modeled a scenario in which AI fully automates research tasks that do not require physical activity. Under that hypothetical, computer science and statistics output could rise as much as 15-fold. In bench sciences such as chemistry and biology, the modeled increase was three to four times, limited by the need for physical work.

Forty percent of surveyed researchers expected AI to conduct entire research projects independently by 2030. The analysis also raised concerns about reduced human oversight, ethical issues and possible effects on researchers’ careers and scientific work. These concerns accompany findings that current Claude use centers on tasks where AI performs well, while idea generation and data gathering remain less common in the sessions examined.