Screen Reader Users in the Vibe Coding Era: Adaptation, Empowerment, and New Accessibility Landscape

Abstract: The rise of generative AI agents has reshaped human-computer interaction and computer-supported cooperative work by shifting users' roles from direct task execution to supervising machine-driven actions, especially in programming (e.g., "vibe coding"). However, there is limited understanding of how screen reader users engage with these systems in practice. To address this gap, we conducted a longitudinal study with 16 screen reader users, exploring their experiences with AI code assistants in daily programming scenarios. Participants first completed a tutorial with GitHub Copilot, then performed a programming task and provided initial feedback. After two weeks of AI-assisted programming, follow-up studies assessed changes in their practices and perceptions. Our findings demonstrate that advanced code assistants not only enhance their programming capabilities but also bridge accessibility gaps. While the assistant proved beneficial, there remains potential to improve how users convey intent and interpret outputs. They also experienced difficulties managing multiple views and maintaining situational awareness. More broadly, they encountered barriers in learning advanced tools and expressed a need to retain control. Based on these insights, we provide design recommendations for more accessible and inclusive AI-assisted tools.