Agentic tools improve web accessibility audits for specific guidelines
Agentic Web Accessibility Auditing: Authoring and Evaluating Per-Criterion Worker Agents for WCAG
Human-Computer Interaction
Summary
Checking websites to make sure they work well for everyone, including people with disabilities, can be hard because automated tools don’t always catch all problems. The authors created a system that uses specialized agents, each focusing on a different accessibility rule from recognized guidelines, to better find issues on websites. Their system was tested on real audits and found more true problems than existing tools, though sometimes with less accuracy. This approach helps auditors see what evidence each agent uses and could make automated website accessibility checking more reliable.
What this means in practice
- •For web developers: Use criterion-specific automated agents to detect accessibility issues matching WCAG guidelines more effectively during website development.
- •For quality assurance teams: Integrate detailed agent-based accessibility checks into testing workflows to improve detection coverage for professional web audits.
Authors
Arjun Mishra, Pranav Karthik, Byungjun Bae, Dongwook Yoon
Abstract
Automated accessibility assessments differ in the evidence they collect and the requirements they address. We present a framework that combines shared browser tools with criterion-specific worker agents, implementing 39 WCAG 2.1 Level A and AA criteria and one additional WCAG 2.2 criterion. We analyze archived predictions on 250 page-criterion records derived from professional audits of scholarly platforms. Workers recover 0.86 of positive reference labels, compared with 0.36 for axe-core and 0.67 for an uncued vision-language model, with lower precision. Criterion-level results, abstentions, development-exposure sensitivities, and separately instrumented runs qualify these comparisons. Inferred negative labels and differences between evaluated configurations limit conclusions about true accuracy and causal effects. We contribute the framework, its criterion-specific implementation, and an evaluation account that distinguishes detection, evidence availability, and resource use, motivating further study of inspectable automated assessments within professional auditing.