Automated LLM-Based Accessibility Remediation: From Conventional Websites to Angular Single-Page Applications

A new research paper proposes a breakthrough in web accessibility by using Large Language Models (LLMs) to automatically fix accessibility issues, moving beyond mere detection. The system, developed by researchers Carla Fernández-Navarro and Francisco Chicano, addresses a critical bottleneck: while tools exist to find accessibility problems, fixing them remains a manual, costly, and error-prone process, especially for dynamic Single-Page Applications (SPAs) like those built with Angular.

The framework presents a modular workflow that actively implements corrections. For static websites, it modifies the live Document Object Model (DOM). For complex Angular projects, it directly patches the TypeScript/HTML source code. This approach overcomes the limitations of traditional static analysis, which struggles with the dynamic nature of modern web apps. In testing, the system successfully fixed 80% of accessibility issues on 12 public static websites and an impressive 86% of issues across 6 open-source Angular applications. A key feature is its ability to generate meaningful alt-text descriptions for images while preserving the application's original design and stability.

This work shifts the paradigm from accessibility as a post-launch audit burden to an integrated part of the development process. By automating remediation, it significantly reduces the time, cost, and expertise required to make websites compliant with standards like WCAG. The research demonstrates that LLMs can be effectively deployed for complex, context-sensitive code generation tasks beyond simple chatbots, offering a scalable solution to a pervasive problem affecting millions of users with disabilities.