AI and automated tools face limits in web accessibility

Artificial intelligence offers significant potential for managing web accessibility at scale, but it cannot replace human oversight. While AI can scan large volumes of digital content, group duplicate findings, and speed up code remediation, it often lacks the judgment required for full compliance. Because frontier AI models are trained on a largely inaccessible web, they risk repeating existing accessibility failures without proper guardrails.

Automated testing tools are limited in their ability to detect all Web Content Accessibility Guidelines (WCAG) violations. Research and practical audits suggest that automated scanners typically catch only between 30 and 57 percent of total failures. These tools are effective at identifying programmatic issues such as insufficient color contrast, missing alt attributes, and improper form labeling. However, they often fail to detect complex usability issues, such as errors that are invisible to screen readers or improper keyboard navigation.

To ensure effective accessibility, experts recommend a hybrid approach: using AI and automated tools to handle high-volume tasks while maintaining human-led manual audits and testing high-impact user journeys with people with disabilities.