A recent report from Anthropic reveals how scientists are primarily leveraging their AI assistant, Claude, for specific research tasks. The study, which analyzed over 200,000 research sessions conducted between August 2nd and September 30th, 2026, found that data analysis and writing are the most common applications.
Key Findings on Claude Usage in Research
The research encompassed 129,000 Claude and Cowork sessions, along with 78,000 Claude Code sessions. Anthropic emphasized that their analysis focused on session classifications like subject area, research sub-field, and specific tasks, ensuring no sensitive research data, hypotheses, experimental methods, preliminary findings, or unpublished papers were extracted or stored. All results were presented in an aggregated format to protect user privacy.
Disciplinary Usage Patterns
The report indicates a notable preference for Claude among researchers in fields like computer science, mathematics, and statistics. Conversely, its use was relatively lower in healthcare and physical sciences that require extensive laboratory work.
Computer Science (CS) emerged as the leading discipline, accounting for 11.9% of all Claude research sessions. Engineering followed at 8.4%, with Medicine at 7.7%. When adjusted for the overall size of research communities within these disciplines, computer science, mathematics, and statistics still showed a significantly higher proportion of Claude usage compared to their representation in the broader research population.
The usage of Claude Code was even more concentrated within quantitative research domains. CS researchers, despite making up only 4.3% of the global research workforce, contributed a substantial 32.7% of Claude Code research sessions. The fields of Artificial Intelligence (AI) and Machine Learning (ML) also showed remarkable engagement. Researchers in these areas, representing just 1.2% of the global research community, accounted for 5% of all Claude research sessions and a significant 15.2% of Claude Code sessions.
In contrast, disciplines such as arts, humanities, and certain social sciences showed comparatively lower adoption rates for Claude. Anthropic attributes these variations not only to the sheer size of research communities but also to how well research tasks align with the capabilities of AI tools like Claude.
Task-Based AI Adoption
When examining specific research tasks, data analysis and writing consistently ranked as the most frequent activities scientists engaged Claude for.
- Data Analysis: This task represented 33.7% of all Claude research sessions and a striking 48.1% of Claude Code sessions. For context, researchers typically dedicate about 27.6% of their work time to data analysis.
- Writing and Communication: These tasks constituted 32% of Claude research sessions, notably higher than the 16% of work time researchers generally allocate to such activities.
Conversely, tasks like data collection, which researchers spend approximately 28.6% of their time on, saw much lower AI engagement, accounting for only 11.5% of Claude research sessions and 11.9% of Claude Code sessions. Research ideation and hypothesis generation tasks had even smaller shares, at 4.2% and 2.1% respectively, compared to the 7.1% of time researchers invest in these areas.
Survey Corroboration
To further validate these findings, Anthropic also conducted a survey involving 528 scientists, predominantly from life and health sciences, who are active users of AI. The survey results largely mirrored the patterns observed in the session data, confirming that data analysis and writing are the primary use cases for AI among these researchers, while data collection, research ideation, and hypothesis generation are utilized to a lesser extent.
