Agentic PR Adoption on GitHub
A project-level study of 25,264 agentic pull requests across 2,361 popular GitHub repositories finds shallow median adoption, stronger relative activity in small projects, and mostly single-human oversight.
TL;DR — A project-level analysis of 25,264 agentic pull requests from 2,361 popular GitHub repositories finds shallow median adoption: one to two agentic PRs per repository over three months. Intensive adoption is concentrated in a small subset, small projects show higher relative activity, and oversight is usually handled by one human reviewer or modifier.
What the study measures
The study examines how agentic coding tools are adopted and managed at the repository level. The abstract motivates the shift by noting that "prior studies have examined PR-level outcomes of agent-generated contributions," while "less is known" about project-level adoption and management.
The reported dataset contains "25,264 agentic PRs from 2,361 popular GitHub repositories." The paper uses that corpus to investigate three stated questions: adoption of agentic coding tools, project-level agentic PR productivity, and human-agent collaboration patterns.
The most important framing point is that the unit of analysis is the project, not just the individual pull request. That makes the findings useful for understanding governance and oversight, not only whether a single agent-generated contribution succeeds.
Full analysis, extracted claims, numerical results, entity graph, FAQ, related work, applications, and BibTeX are available via x402 micropayment.