Skip to content
UX Atlas
AI workflowCraft conventionAdvanced

UI to Accessibility Audit

Using a model alongside automated tooling, which extends coverage a little and does not approach a real audit.

Definition

A model can review markup or a screenshot against WCAG criteria and flag likely failures. Combined with an automated tool it finds more than either alone, and both together still miss most manual criteria.

Craft convention. True because the industry converged on it. Breaking it costs familiarity, not correctness.

On this page
  1. Definition
  2. What each method actually covers
  3. Prompt for a markup review
  4. Continue from here

What each method actually covers

MethodCoversMisses
Automated toolsContrast, missing labels, ARIA misuse, heading orderAnything requiring judgement or interaction.
Model on markupSemantics, label quality, heading logic, likely ARIA errorsRendered output, focus order, dynamic behaviour.
Model on screenshotVisual contrast, target size, colour-only signalsEverything not visible in a static image.
Manual keyboard and screen reader testingMost of what mattersNothing, but it costs time.

Prompt for a markup review

text
Review this markup against WCAG 2.2 level AA.

For each likely failure return:
- The success criterion number and name
- The specific element
- Why it fails
- The corrected markup

Then list separately the criteria you CANNOT assess from markup alone
and what manual test each requires.

Do not report a failure unless the markup shows it.

Markup:
[paste]

Continue from here

Each link says what the connection is, so you can tell a principle from an alternative from a thing people mix this up with.

Often used with

These usually appear in the same screen or the same decision.

Check your work

Review passes that test whether this was done properly.

Related concept

Connected closely enough to change how you apply this.