Generative AI for Developers with Disabilities

Generative AI tools have become increasingly integrated into software development workflows and are frequently associated with measurable improvements in developer productivity. However, while existing research has primarily focused on the general developer population, limited attention has been given to how the effects of AI-assisted development tools extend to developers with disabilities. This study investigates whether the wider availability of Generative AI tools was associated with changes in observable productivity-related activity among GitHub users who self-identify as having a disability. From an initial pool of over 3000 users, we curated a dataset of 199 GitHub users across multiple disability categories, using publicly available profile information. We separately identified explicit references to AI systems in commit messages as indicators of documented AI use. Productivity was evaluated within and across disability groups using metrics such as commits, code contributions, and pull request acceptance rates, which are commonly used as proxies for developer productivity. We found limited evidence of measurable productivity-related changes following the wider availability of Generative AI tools. While most disability groups did not exhibit statistically significant differences across the chosen metrics, the group identifying themselves as disabled showed a statistically significant increase in net lines modified after the wider availability of Generative AI tool. This suggests that the relationship between the wider availability of Generative AI tool and developer productivity may be more complex than expected, and that GitHub metrics alone may not adequately capture developers' experience.