AI Prompts Shared for WCAG 2.2 Audits
A set of AI prompts designed to automate WCAG 2.2 AA accessibility audits was shared on social media. The prompts provide a structured approach covering perceivable, operable, understandable, and robust criteria, along with mobile and cognitive checklists. The tool aims to streamline the identification of accessibility issues and suggest remediation steps.
- Automated accessibility testing tools, which often use AI, can only detect an estimated 20-30% of all WCAG issues, as they struggle with nuances requiring human judgment. For example, an AI can check if an image has alternative text, but it cannot determine if that text is meaningful or accurately conveys the image's context. - The European Union's harmonised standard for digital accessibility, EN 301 549, is expected to incorporate WCAG 2.2 in a new version (4.1.1) with a publication planned for early 2026. This update will align the standard with the latest accessibility guidelines for public procurement of ICT products and services. - For public sector bodies in the EU, the Web Accessibility Directive already mandates compliance with WCAG standards for websites and mobile apps. The European Accessibility Act, with enforcement beginning June 28, 2025, extends similar accessibility requirements to a range of private sector products and services. - AI prompts for accessibility can be designed to perform specific tasks such as generating a developer handoff checklist for WCAG 2.2, creating a VoiceOver test script for a critical user path, or reviewing UI components for accessibility gaps. Other prompts can analyze HTML for a logical heading hierarchy or check color pairs for sufficient contrast ratios. - A significant challenge in relying on AI for accessibility audits is the potential for false positives and negatives, where the tool might incorrectly flag a compliant element or miss a genuine issue. This is because automated checks often cannot understand the context of the content or the overall user experience. - The new success criteria in WCAG 2.2 particularly address usability on touch-screen devices, cognitive accessibility, and the visibility of focus indicators. For instance, the "Focus Not Obscured" criterion ensures that elements like cookie banners do not hide the item a keyboard user is focused on. - While there are examples of AI being used in the European public sector for services like virtual assistants and data analysis, the adoption of AI specifically for comprehensive accessibility audits is not yet widespread. Barriers to adoption include the need for significant investment in organizational capital and the slow adaptation of legacy IT systems. - Cautionary examples from the public sector, such as the Dutch childcare benefit scandal where an AI system used for fraud detection led to discrimination, highlight the risks of deploying AI without sufficient human oversight and inclusive design. This underscores the importance of a human-centered approach when using AI for accessibility evaluations.
