Which accessibility checks need expert review?

Expert review is essential for accessibility checks that depend on human judgement: whether content makes sense, whether keyboard and screen-reader journeys work, whether AI output is accurate, and whether design choices exclude people in real situations. Automated tools can help find repeatable defects, but they cannot replace empathy, context or lived experience.

Accessibility evaluations require specialist assessment where context shapes access

To know whether manual accessibility testing is needed, consider whether a check depends on how a person understands, navigates or experiences a digital service. Expert reviewers should assess whether content works for people using keyboards instead of a mouse, browsers that read pages aloud, different devices and different connection speeds. They should also look beyond permanent disabilities. Bright sunlight, distraction, limited time or reading in a second language can all change what makes a website usable.

The key test is simple: does the experience work for the person trying to complete a task, not merely for the code being scanned? Automated accessibility scanning can flag patterns, but it cannot reliably judge whether a journey is clear, whether an interaction is frustrating or whether a design decision creates an avoidable barrier.

Sources: Designing and developing web experiences for everyone, The Prompt: Using AI to Create a More Accessible and Dynamic Digital Future

Review AI fixes and overlays before they create a false sense of progress

AI-generated accessibility content, automated fixes and accessibility overlays all need expert review because accuracy is not guaranteed. AI can misinterpret an image, miss cultural context or reproduce bias, making a seemingly helpful output less inclusive in practice. Human designers, developers and accessibility advocates need to check the result against the real experience.

Overlays need particular caution. A toolbar added on top of a website may address a small share of WCAG failures while leaving the underlying structure untouched. That can frustrate users, affect performance and create legal risk without solving the barrier. The stronger approach is accessible-by-design work in the architecture, content, interaction patterns and code, followed by human evaluation of critical user journeys.

Sources: The Prompt: Using AI to Create a More Accessible and Dynamic Digital Future

We believe accessibility needs human empathy as well as technical checks

We believe expert review should focus on the moments where a tool cannot understand a person: reading an image in context, navigating without a mouse, following a screen-reader journey or making sense of content under pressure. That is where accessibility becomes a human experience, not a score. We also believe quick fixes are rarely enough. Our accessibility writing has challenged AI overlays that sit on top of broken structures and has made the case for keeping designers, developers and accessibility advocates involved when AI is used. We see AI as a useful collaborator for repetitive work, but not as a replacement for the care, context and empathy that inclusive digital experiences require.

Sources: The Prompt: Using AI to Create a More Accessible and Dynamic Digital Future, Designing and developing web experiences for everyone

FAQs

Do keyboard accessibility checks need expert review?

Yes. Keyboard accessibility needs expert review because people may navigate without a mouse, and a reviewer must judge whether the resulting journey is practical and understandable. The review should consider the full task, not only whether an element can technically receive focus.

Does screen-reader testing need human review?

Yes. Screen-reader testing needs people to assess whether a page communicates meaning clearly when content is read aloud and navigation happens through assistive technology. A technically present element is not enough if the experience remains confusing or incomplete for the person using it.

Can AI write accessibility content without human review?

No. AI can help scale repetitive accessibility work, but human review remains necessary because AI may misinterpret images, misunderstand cultural context or reflect bias. Designers, developers and accessibility advocates should remain responsible for whether the output works for people.

Are accessibility overlays a substitute for expert testing?

No. Accessibility overlays may address only a limited share of WCAG failures and do not correct the underlying structure where accessibility problems often sit. Expert review is still needed to assess the real experience and the root cause of barriers.