For a comprehensive accessibility audit, several thousand pages in a client project needed to be technically analyzed and also evaluated qualitatively. While a UX colleague worked with screen readers, keyboard navigation, and manual testing, my task was to use automation to identify as many violations of the WCAG guidelines as possible. Although there were numerous tools available for individual aspects of the audit, a consistent, end-to-end process was lacking. Technical testing, qualitative evaluations, documentation, and ticket management largely ran in parallel. This meant that all results had to be manually consolidated and evaluated individually.

Milestone: A First Prototype

This situation led to the creation of an initial technical prototype. The idea was simple: automatically crawl a website, capture all accessible URLs, and then analyze them using an accessibility testing engine. For the initial task at hand, this was already a major step forward. Instead of manually checking individual pages, entire domains could now be examined automatically. At the same time, the limitations of this approach quickly became apparent. The solution was functional, but still a long way from being a real product. The evaluation was based on extensive JSON files; there was no user-friendly front end, and, most importantly, only static pages could be checked. And the biggest catch: modern websites, in particular, have long since ceased to consist solely of individual pages. Forms, rate calculators, configurators, and other interactive applications often play a central role and are particularly relevant from an accessibility perspective, but they can hardly be tested automatically using the original approach.

What does the ideal accessibility tool look like?

An ideal accessibility tool should combine technical and qualitative results, enable automated testing, and integrate identified violations directly into existing development processes. When examining whether existing platforms can meet these requirements, the results are sobering. None of the solutions tested could truly meet the requirements. The tools quickly reached their limits, especially with more complex user flows.

The ideal accessibility tool: an automated approach using AI

Advances in AI are opening up new possibilities for automated testing. When AI is used as a planner, the system analyzes the structure of an application and independently generates a test plan based on that analysis.

The advantages of an AI-powered accessibility tool

• Interactive elements such as forms, buttons, or modal windows are recognized and combined into realistic usage scenarios.

• Browser automation handles the actual execution of these steps.

• After each action, the system checks again to see if any accessibility issues arise.

• Any violations found can be transferred directly into Jira tickets, along with details on the affected elements, relevant metadata, and information about the specific WCAG violation.

• This approach makes it possible, for the first time, to systematically examine dynamic states and complex interactions.

It’s important to remember: Despite all the progress made, accessibility cannot be fully automated. Technical checks are excellent for quickly identifying large numbers of issues. They analyze code, test interactive components, and reliably detect recurring errors. However, they do not replace the human perspective. Screen reader usage, keyboard navigation, content comprehensibility, and cognitive accessibility still require qualitative assessments by experienced experts and real users. That is precisely why we do not view this tool as a replacement for accessibility audits, but rather as a complement to them. True quality, however, continues to come from a combination of automation, UX expertise, and continuous improvement.