GitHub Copilot increases code contributions in open source projects by 5.9 percent at the project level, while simultaneously raising coordination time by eight percent. This is shown in a study by Fangchen Song, Ashish Agarwal, and Wen Wen, published in its fourth version on arXiv on August 13, 2026.
The researchers combined proprietary GitHub Copilot usage data with public project data from GitHub and analyzed the impact of the AI pair programmer on collaborative development in open source software projects.
Productivity gains across multiple metrics
According to the study, Copilot increased individual code contributions by 2.1 percent and developer coding participation by 3.4 percent. The average number of pull requests per developer grew by 5.5 percent.
These figures are based on longitudinal analyses across multiple project versions. The authors interpret the results as evidence that generative AI increases developers' ability to produce code and expands participation in open source projects.
Coordination costs rise in parallel
The study documents an eight percent increase in coordination time. Code discussions also intensified. The researchers identify this tension as a central finding: while GitHub Copilot expands who can contribute and how much is contributed, the tool simultaneously slows down coordination in collective development efforts.
Despite increased coordination time, the overall effect remains positive. The study records a net increase in timely merging of code contributions at the project level. Productivity gains thus outweigh coordination losses.
Differential impact by developer role
A central finding of the research concerns heterogeneous effects by position within the project. Peripheral developers show relatively smaller increases in project-level code contributions but larger increases in coordination time compared to core developers.
According to the study, core developers benefit more strongly from productivity gains with lower coordination costs. This heterogeneity suggests that GitHub Copilot affects different groups within the open source community differently.
Methodological approach and data foundation
The study is based on a combination of proprietary GitHub Copilot usage data and publicly available OSS project data from GitHub. The research was originally submitted on October 2, 2024, and revised through its fourth version on August 13, 2026.
The authors emphasize that their results highlight the dual effects of AI programming tools on code contributions and coordination in open source development. The findings provide insights into how generative AI could reshape the structure of OSS communities over time. The dynamic between centrally acting core developers and peripheral contributors may shift further.