AccessGuru: Leveraging LLMs to Detect and Correct Web Accessibility Violations in HTML Code

2025-07-24

AI Summary

Web accessibility violations are pervasive, yet existing automated detection and repair techniques suffer from limited coverage and suboptimal efficacy. This paper proposes a novel large language model (LLM)-driven framework: first, we construct a fine-grained taxonomy of accessibility violations across three dimensions—syntax, semantics, and layout; second, we design a taxonomy-aware prompting strategy that synergistically integrates LLMs with conventional detection tools to enable unified violation detection and end-to-end repair; third, we employ human-in-the-loop verification and comparative evaluation to rigorously ensure repair correctness and quality. Evaluated on a real-world web benchmark, our approach reduces average violation scores by 84%, substantially outperforming the prior state-of-the-art (50%). To the best of our knowledge, this is the first fully automated framework capable of jointly identifying and repairing violations across all three categories. It offers a scalable, interpretable, and practically deployable pathway toward enhancing Web inclusivity.

Technology Category

• Humans and AI: AI for Accessibility
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

The vast majority of Web pages fail to comply with established Web accessibility guidelines, excluding a range of users with diverse abilities from interacting with their content. Making Web pages accessible to all users requires dedicated expertise and additional manual efforts from Web page providers. To lower their efforts and promote inclusiveness, we aim to automatically detect and correct Web accessibility violations in HTML code. While previous work has made progress in detecting certain types of accessibility violations, the problem of automatically detecting and correcting accessibility violations remains an open challenge that we address. We introduce a novel taxonomy classifying Web accessibility violations into three key categories - Syntactic, Semantic, and Layout. This taxonomy provides a structured foundation for developing our detection and correction method and redefining evaluation metrics. We propose a novel method, AccessGuru, which combines existing accessibility testing tools and Large Language Models (LLMs) to detect violations and applies taxonomy-driven prompting strategies to correct all three categories. To evaluate these capabilities, we develop a benchmark of real-world Web accessibility violations. Our benchmark quantifies syntactic and layout compliance and judges semantic accuracy through comparative analysis with human expert corrections. Evaluation against our benchmark shows that AccessGuru achieves up to 84% average violation score decrease, significantly outperforming prior methods that achieve at most 50%.

Problem

• Automatically detect and correct web accessibility violations in HTML
• Classify accessibility violations into Syntactic, Semantic, and Layout categories
• Improve compliance using LLMs and taxonomy-driven prompting strategies

Innovation

• Combines LLMs and testing tools for accessibility
• Introduces taxonomy for three violation categories
• Uses benchmark for real-world violation evaluation