How AI Will Transform Accessibility Testing

It’s no secret that digital accessibility experts today are pressed for time and resources. This is particularly true when it comes to testing already-built digital experiences. Our current approach to testing is to leverage both automation and manual review. However, manual testing can be slow, and it requires specialized knowledge—making it difficult for experts to share the work, even as the velocity of digital content creation continues to increase. And while automated testing can accelerate progress, what can be detected by traditional automated testing is nearing its limits.

Proactively creating inclusive experiences from the start and applying automated remediation (automatically fixing for the end user common accessibility errors) can reduce the need for extensive testing. However, validation testing (testing after remediation has taken place to ensure issues have, in fact, been resolved) is still essential.

Enter artificial intelligence (AI). AI is rapidly changing the digital accessibility testing landscape, promising to help experts and the organizations they serve overcome the challenges posed by our current approach. Not only can AI improve the efficiency of testing, but enhancements in the accuracy of AI-powered automated testing tools allow more people—including non-experts—to identify accessibility barriers, opening the door to broader participation in testing efforts.

In the last installment of my series of AI and digital accessibility, I’ll explore the ways in which AI will accelerate and democratize accessibility testing, driving progress toward a more inclusive digital world.

(If you’re new to this series, consider starting with part one, The Impact of AI on People with Disabilities, and part two, AI and Accessibility in the Digital Experience Life Cycle.)

AI component detection will build on efficiency gains.

Before evaluating whether a specific user flow—the set of steps a user takes to complete a core task, like making a purchase—is accessible, testers must first determine the right tests and accessibility patterns to apply. These decisions are made based on what types of components make up that user flow. For example, if a user flow contains page tabs, tests for specific keyboard interactions (like arrow keys and “enter” and “space” commands) may be applied.

Traditionally, this work has been performed manually by experts, and it can be time-consuming. However, the same user interface (UI) components are often used throughout an experience—and for some time, AI has been used to visually detect common components. As AI advances, AI tools equipped with visual component detection will be able to even more effectively identify common UI elements and automatically apply an increasing number of relevant tests and patterns.

For example, if a page tab interface is detected, AI tools would trigger tests for required properties and keyboard accessibility patterns. These AI capabilities will ultimately result in faster, and more scalable, testing, enabling teams to more easily understand which issues are repeated in components used across their digital properties and should be prioritized.

Accuracy enhancements through AI issue detection will help democratize testing processes.

At present, automated tests are limited in their capability and accuracy. These limitations are the primary reason organizations must rely on experts with specialized knowledge to manually evaluate digital experiences. But AI has the potential to significantly improve the accuracy of automated testing, empowering content creators—such as web and document authors—who are not accessibility experts to more easily test the accessibility of their work.

Training testing tools with accessibility data that has been validated by experts will enhance these tools’ ability to detect a broad range of issues. These advancements will enable teams to confidently automate more testing, reducing the need for manual evaluation by experts.

Some people may be skeptical that AI will be capable of performing tests that are as comprehensive and reliable as those conducted manually by experts today. However, foundational models, including large language models (LLMs) like Chat GPT, can already effectively test various Web Content Accessibility Guidelines (WCAG) criteria. These models can, for example, ascertain the language of a web page or text passage and verify the correct language attributes in the code.

Among other things, LLMs are also capable of evaluating link text within its context—matching the accessible name of a link to its contextual presentation within paragraphs, sentences, list items, or table cells. And these models can assess whether section headings accurately reflect the content that follows. I predict it’s only a matter of time before AI-powered automated testing tools become sophisticated enough to perform accessibility evaluations that are comparable in scope and accuracy to those performed by experts, empowering many more people to test the accessibility of digital content.

AI remediation guidance will streamline fixes and reduce knowledge barriers.

Teams today don’t just need specialized knowledge to perform accessibility testing. Often, expert support is required to correct the issues that tests surface. I believe that AI testing tools will soon be able to provide teams with clear, simple guidance for resolving accessibility barriers—meaning individuals won’t need to be experts in specific programming languages or UI frameworks to make improvements to their digital experiences’ accessibility.

Experienced developers will also benefit from AI-powered remediation guidance, particularly when it comes to resolving complex accessibility barriers. Access to detailed recommendations for how to fix issues will save development teams time, resulting in faster enhancements to users’ experience.

Eventually, the remediation process itself may become largely automated. In my last article, I discussed how automated remediation tools can already find, and fix, many common accessibility barriers for users, while developers work on more permanent code-level solutions. And in the future, advanced AI tools may be capable of applying fixes directly in the experience's source code. Developers would need only to approve remediation recommendations made by these tools.

This shift would drastically streamline one of the most challenging and tedious aspects of ensuring digital experiences are accessible, resulting in better experiences for users with disabilities. Too often, audits only create a longer list of tickets sitting in developers’ backlogs. As AI becomes more integrated into issue identification and remediation, more barriers can be immediately addressed for users.? ??

AI-powered testing solutions will allow for new accessibility standards that elevate user experience.

Beyond saving teams time and democratizing testing and remediation, AI could ultimately elevate the status-quo for accessible digital content. By making testing easier for teams, AI-based automated testing solutions could lead to new accessibility guidelines with greater benefits for users with disabilities.

Some current accessibility guidelines are limited in scope, accounting only for what can be practically tested. They may not require certain outcomes if testing for a specific requirement is too complex. For example, testing for sufficiently visible focus indicators can be time-consuming, as it involves calculating the area that makes up the focus indicator. As a result, the adoption of some criteria is limited.

AI-based automated solutions could make testing for visual focus indicators, and other accessibility outcomes, far less challenging. This would allow for new guidelines that raise the bar for accessible content without increasing the amount of testing time needed to meet these requirements.

Accelerating progress toward an inclusive digital future

By allowing teams to shift more responsibilities to intelligent systems, advancements in AI have the potential to improve the accuracy of accessibility testing, ultimately paving the way for a more inclusive digital future. Teams will no longer have to spend valuable time on repetitive testing, and can focus instead on problems that require human expertise. Most importantly, with more accurate testing, teams can devote time to proactively incorporating accessibility best practices in other phases of the experience creation life cycle—more quickly and efficiently delivering barrier-free experiences for all users.