Where Do AI Coding Agents Fail? An Empirical Study of Failed Agentic Pull Requests in GitHub
Ramtin Ehsani, Sakshi Pathak, Shriya Rawal, Abdullah Al Mujahid, Mia Mohammad Imran, Preetha Chatterjee
TL;DR
The paper investigates why AI coding agents fail to have their autonomous pull requests merged in real-world GitHub projects. It conducts a large-scale empirical study of 33k agent-authored PRs from five coding agents using four quantitative axes (task type, code changes, CI outcomes, review dynamics) and complements this with a qualitative analysis of 600 rejected PRs to derive a hierarchical rejection taxonomy. Key findings show higher merge rates for documentation, CI, and build tasks, while performance and bug-fix tasks lag; not-merged PRs tend to feature larger, more invasive changes and more CI failures, with reviewer abandonment emerging as the dominant rejection pattern alongside duplicates and unwanted features. The study provides practical guidance for designing more context-aware, collaboration-sensitive AI coding agents and for improving the integration of autonomous contributors into real-world software development workflows.
Abstract
AI coding agents are now submitting pull requests (PRs) to software projects, acting not just as assistants but as autonomous contributors. As these agentic contributions are rapidly increasing across real repositories, little is known about how they behave in practice and why many of them fail to be merged. In this paper, we conduct a large-scale study of 33k agent-authored PRs made by five coding agents across GitHub. (RQ1) We first quantitatively characterize merged and not-merged PRs along four broad dimensions: 1) merge outcomes across task types, 2) code changes, 3) CI build results, and 4) review dynamics. We observe that tasks related to documentation, CI, and build update achieve the highest merge success, whereas performance and bug-fix tasks perform the worst. Not-merged PRs tend to involve larger code changes, touch more files, and often do not pass the project's CI/CD pipeline validation. (RQ2) To further investigate why some agentic PRs are not merged, we qualitatively analyze 600 PRs to derive a hierarchical taxonomy of rejection patterns. This analysis complements the quantitative findings in RQ1 by uncovering rejection reasons not captured by quantitative metrics, including lack of meaningful reviewer engagement, duplicate PRs, unwanted feature implementations, and agent misalignment. Together, our findings highlight key socio-technical and human-AI collaboration factors that are critical to improving the success of future agentic workflows.