What is AI accessibility testing?
Classic automated accessibility tools scan a page's code for rule violations, such as an image with no alt attribute or text with low contrast. They are fast and useful, but they only see one page state at a time and only what a rule can express.
AI accessibility testing adds an agent that uses the product the way a keyboard or screen reader user would. It tabs through a checkout, opens menus, submits forms with errors, and checks whether focus, names and messages make sense along the way. It can also compare what it finds with your design system and requirements.
In ShipperAG this is the job of the accessibility specialist, one of the AI QA agents, which works alongside the user journeys, visual design and personas specialists.
WCAG 2.2 in brief
The Web Content Accessibility Guidelines are the W3C standard most accessibility laws and policies point to. WCAG 2.2 is a W3C Recommendation and adds nine success criteria to WCAG 2.1, including:
- Focus Not Obscured (Minimum): a focused element should not be hidden behind sticky headers or banners.
- Target Size (Minimum): controls need to be large enough, or spaced enough, to hit reliably.
- Dragging Movements: anything done by dragging needs a single-pointer alternative.
- Consistent Help and Redundant Entry: help stays in a predictable place, and users are not asked for the same information twice.
- Accessible Authentication (Minimum): logging in should not depend on a cognitive test like remembering or transcribing a code without support.
Several of these are about behaviour across a flow rather than a single page, which is where an agent that actually walks through journeys helps. To work through all 55 Level A and AA criteria yourself, try our free WCAG 2.2 quick check, an in-browser checklist with an instant summary.
The European Accessibility Act
The European Accessibility Act (Directive (EU) 2019/882) sets accessibility requirements for a range of products and services in the EU. According to the directive on EUR-Lex, Member States apply its measures from 28 June 2025, and covered services provided to consumers after that date must meet its requirements.
The European Commission's EAA overview lists e-commerce among the covered services, alongside banking services, e-books, transport services and others. For many online shops and the product teams that build them, accessibility is no longer only good practice.
The harmonised standard behind the EAA, EN 301 549, was updated to version 4.1.1 in September 2026 and aligns with WCAG 2.2, although it has not yet been cited in the Official Journal. For scope, dates, the micro-enterprise exemption and a practical testing checklist, read our European Accessibility Act testing guide.
This page is general information, not legal advice. Check your own obligations with a qualified adviser.
What automated accessibility checks can and cannot do
This is where honesty matters most. The W3C's own guidance on evaluation tools puts it plainly: "Tools cannot check all accessibility aspects automatically." (W3C WAI). Human judgment is still required.
Good candidates for automated and AI checks
- Every interactive control reachable and usable by keyboard.
- Visible focus that is not hidden behind other content.
- Form fields with labels, and error messages tied to the right field.
- Text and interface contrast against your actual design tokens.
- Target sizes, accessible names and roles, and heading structure.
- Regressions: a component that was accessible last release and is not now.
What still needs human review
- Whether alt text and link text are actually meaningful in context.
- Caption and transcript quality, and reading order that makes sense.
- Cognitive load, plain language and whether a flow is understandable.
- Real experience with assistive technology, ideally tested by disabled users.
A tool that reports "100% accessible" is overclaiming. ShipperAG reports what was verified and lists the rest as not covered or needs input, so your team knows exactly where human review should focus.
How ShipperAG approaches AI accessibility testing
- Context first. It reads your design tokens and components, so contrast and focus checks use your real colours and states.
- Journeys, not just pages. It walks key flows like sign-up and checkout by keyboard, including error states.
- Verified findings. An independent verifier re-runs each issue before it is reported, with screenshots and steps.
- Every release. Checks run on each change, so accessibility does not silently regress between audits.
- Evidence you can share. Results go into an evidence ledger and release report.
For teams selling to consumers in the EU, that report can become part of release sign-off. For how verification works across all agents, read agentic QA testing explained.