An Assessment of LLM-Based Auditing and Validation for Web Accessibility

Accessibility constitutes a fundamental right codified in national and international legislation and the SDGs. In particular, web accessibility is essential to enable people with disabilities to engage fully in societal and professional activities. Despite mandatory legal requirements, such as those included in the Web Accessibility Directive and the European Accessibility Act, a substantial portion of websites continue to exhibit accessibility barriers. Moreover, awareness of accessibility best practices remains limited among designers and developers. In this context, we assess the effectiveness of Large Language Models for automated accessibility auditing and HTML validation. We also examine how these models can aid developers in producing semantically correct and accessible web content, and we discuss their current limitations. Our research highlights the promises and challenges of integrating LLMs in web development and advocates using these tools as supportive aids rather than standalone solutions.