Designing and Evaluating an AI-Powered Figma Plugin for Accessibility Education

Abstract

Digital accessibility has received increasing attention in computing and design education, with a growing body of work advocating for its inclusion in curricula. However, much of this emphasis remains concentrated in later stages of software development, such as code-level auditing and compliance checking, with limited integration into early design workflows. As a result, the shift-left principle, which addresses accessibility from the initial UX design phase, remains insufficiently implemented in practice. To address this gap, we developed a Figma plugin that uses a multimodal LLM that evaluates UI frames against WCAG criteria and provides contextual, design-stage feedback to support early identification and remediation of accessibility issues. In a three-week deployment, 89 practicing professionals and computing or design students used the plugin alongside existing tools (Stark, Axe for Designers) and they reported their perceptions, experiences, and self-reported abilities before and after use through a pre-post survey. Findings show that 92.1% of participants reported catching accessibility issues earlier in their workflow, 96.6% reported productivity gains, and recommendation likelihood was high (M=4.25/5), while generated specifications were rated accurate (M=4.08/5). Findings also indicate that integrating AI-supported feedback within the design environment encouraged participants to consider accessibility earlier in their workflow and improved their ability to identify and reason about accessibility issues.