How generative AI is shaping research software development and maintenance at a research-intensive university

Generative artificial intelligence is spreading rapidly across academic research, yet its role in the development and maintenance of research software remains insufficiently characterized.

A six week, institutional review board approved, anonymized online survey of faculty and research staff at a large research intensive university in late 2024 (n = 251). Branching survey questions distinguished general users of research software from those who create or maintain it. Quantitative associations were examined using chi square or Fisher's exact tests, and free text descriptions of generative AI use in software development were analyzed thematically.

Overall, 29% of respondents reported using generative AI for at least one research task. Within the subsample of active research software developers, 33% reported using generative AI for software development and 51% indicated continued or planned future use. No statistically significant associations were found for age, recency of highest degree, or external funding. Gender was significantly associated with generative AI use for software development, with higher uptake among men than women (41% versus 15%; χ ²(1)=5.03, p=.025). Reported generative AI uses clustered around four practical roles: generating initial code and queries, supporting debugging and testing, transforming data or commands via natural language prompts, and reducing cognitive burden in repetitive or complex tasks.

At a large research intensive university, generative AI adoption in research software development is already common among active developers and is expected to expand. The observed gender disparity signals a potential equity risk as tool assisted development becomes normalized. These findings provide an empirical baseline for multi institution replication and for evaluating how generative AI may reshape the organization and distribution of research software work.