This post is based on research for a course I developed for the Vispero Tutor training platform. The course is part of a role-based curriculum on using artificial intelligence (AI) tools safely and responsibly to support digital accessibility. Other courses in the curriculum focus on designers, developers, content authors, and a foundation course for digital practitioners.
This article is intended as a brief summary of some key points on how accessibility auditors and testers can use AI.
The Unique Role of Accessibility Auditors
Accessibility auditors and testers occupy a particular position in the accessibility lifecycle.
Product managers, designers, developers, content authors, and QA testers can legitimately use AI to explore possibilities during the product development process.
Auditors carry a different evidentiary burden. If an auditor reports a failure, assigns a WCAG success criterion, describes user impact, recommends remediation, or says something passes, that conclusion must be explainable and defensible.
Other people may make important decisions based on an auditor’s evaluations.
- A developer may change code because of an audit finding.
- A product owner may delay a release.
- An organization may make a conformance claim.
- A procurement team may accept or reject a product.
- A customer may rely on an accessibility report.
Website owners and product development team members may challenge the results of an audit, claim that an issue is insignificant or irrelevant, request more detail, and/or ask for a different solution.
That means accessibility testing must fundamentally be an evidence-based activity.
AI can play a useful role in establishing the evidence for an auditing report and make auditors significantly more productive.
However, an AI system or tool must not become the sole source of an audit conclusion.
AI can strengthen accessibility evaluation by supporting the professional judgement and accountability of the auditor, on which trustworthy auditing depends.
AI Reality
Accessibility discussions about AI in general often fall into one of two extreme camps.
- AI is the future and will solve accessibility for people with disabilities.
- AI is biased, unreliable, and harmful to people with disabilities.
In my opinion, the reality is much more subtle and interesting. AI is neither a magic solution nor something to be dismissed.
It’s a tool, an unusually powerful one, that can help accessibility professionals work more effectively in several specific ways. It offers unprecedented speed, scale and informational reach.
As a small example, the W3C’s WCAG Evaluation Methodology is based on testing a representative sample of digital content that may be limited to a handful of key pages, screens and user workflows. Used correctly, AI can analyse thousands of web pages in minutes.
But it still needs expertise, purposeful testing, critical thinking, and human judgement in order to deliver meaningful outcomes, as well as specific mechanisms to ensure the voices of people with disabilities are heard.
Auditors must remember that it is their expertise and skills that will inform their reports and that they will carry the responsibility for accuracy, not the AI tools.
AI should be regarded as an auditor’s partner – and a junior partner, at that. Used correctly, it can become an important part of an auditor’s toolkit.
AI Tools
AI can strengthen the various ways of testing accessibility that we already use.
Automated Testing
While automated tools are limited in how many issues they can identify – typically 30%-60%, they are still good first-pass tools, saving auditors time they can spend on more detailed manual testing.
Many automated testing tools now make use of AI to improve their results, which can speed up the process and cover more territory, but also necessarily requires human review.
Manual Testing
Manual testing is required to identify issues missed by automated testing tools, whether they’re AI-powered or not.
Many experienced auditors prefer to skip automated testing and go straight to manual testing, which typically includes three major activities:
- Reviewing Content: Generative AI in the form of conversational chatbots using Large Language Models and Machine Learning is a useful tool for reviewing planned or existing digital content. The key is to learn how to prompt these models, question the results, and seek verification.
- Navigating with Keyboard Alone: AI tools can simulate tabbing, flag coding errors, and analyse layout flow. What they can’t do reliably is detect complex keyboard traps, evaluate whether focus indicators are sufficiently visible, or understand the real world frustration of an inaccessible user experience.
- Using Assistive Technology: AI is becoming increasingly embedded into AT, extending and improving its ability to convey digital content to users with disabilities. At a minimum, an accessibility auditor will use a screen reader as part of their manual testing. When an auditor is testing digital content with a screen reader they should ideally test with AI tools enabled and also without, to reflect the reality that not every user will have AI tools installed or enabled.
Real Users
Any good auditor knows that the acid test for digital accessibility is usability testing with real users with disabilities.
You may be limited in how much of this you can organize, especially given the range of disabilities that people have and the logistical challenges involved, but invariably this kind of testing reveals issues that are simply not identified in any other way.
While you may believe yourself to be adept navigating, understanding, and operating digital in a testing environment, you’ll find that real users with disabilities do it in completely different ways, whether AI is involved or not.
And, while AI will offer to test with personas that simulate real users, the results are not currently – and may never be – either convincing or reliable.
AI Value
The highest-value role for AI is not finding obvious WCAG failures; that’s what automated tools already do well.
The highest value lies in helping expert auditors think more broadly, cover more disability perspectives, explore edge cases, improve consistency, and reduce the administrative burden of auditing.
Used this way, AI can make a strong auditor more effective without replacing the judgement, experience, and accountability that accessibility auditing requires.
Use AI as a two-way street in your workflow, checking its work and having it check yours.
You don’t need to become an AI expert to benefit from AI in digital accessibility auditing. Start with the accessibility expertise you already have. Use AI to extend that expertise, not replace it.
If you do that, you’ll be well positioned to use AI effectively, responsibly, and practically from day one.
Be curious, be sceptical, verify what matters, and keep people with disabilities at the centre of every auditing decision.