What Can A11yCheck Review Automatically?

Questions answered by this guide

• ?Can an automated scan detect every accessibility issue?
• ?What does A11yCheck check automatically?
• ?What still needs a human reviewer?
• ?Does A11yCheck use AI, and how reliable is it?
• ?Can I trust an automated accessibility score without a human review?
• ?What's the difference between an accessibility finding and a course quality finding?

What you'll learn

• A11yCheck's scan reliably detects structural, technical, and measurable issues
• It can't judge instructional quality, learning objectives, or subjective meaning
• Every issue-specific Learning Hub guide states this split for its own topic; this page is the complete picture in one place

What's inside

Instructional designers, L&D managers, and QA reviewers deciding how much to trust an automated scan

1. 01What gets checked automatically
2. 02What still needs a human
3. 03Why the distinction matters
4. 04How this fits into a real review

Why it matters

An automated scan is fast, consistent, and never gets tired halfway through a 40-slide course, but it can only check what's actually measurable: does this image have alt text, does this text meet a contrast ratio, can this button be reached by keyboard. It cannot judge whether an image's alt text accurately describes what's in the image, whether a learning objective is well written, or whether an interaction genuinely helps someone learn. Being clear about that line, in public, is what makes an automated tool trustworthy instead of just fast.

When you'll encounter this

• You're deciding whether a clean A11yCheck scan means a course is fully ready to publish.
• A stakeholder asks whether an automated tool can replace a human accessibility review entirely.
• You're comparing A11yCheck's Accessibility, Course Quality, and Course Engagement findings and want to know how confident each one is.
• You're evaluating whether to trust an AI-assisted scanning tool at all.

Example

Best practices

• Missing or empty alt text on images, icons, and graphics.
• Low color contrast between text and its background, and on icons or UI components.
• Elements that can't be reached or activated using a keyboard alone, and tab order that doesn't match a slide's visual layout.
• Missing captions or transcripts on narrated slides and video.
• Interactive targets smaller than the accessible minimum, or crammed too close together.
• Text that gets cut off or overlaps when a learner zooms in.
• Data tables with no real header structure, and headings that are styled text instead of real headings.
• Broken links, inconsistent styling, and duplicate slide or lesson titles (Course Quality).
• Slides with more text than a learner can reasonably read in the time given (Course Engagement, advisory, not pass or fail).

Common mistakes

• Treating a 100% or "clean" scan result as proof a course is fully accessible to every learner, rather than proof it passed every check a machine can measure.
• Skipping a manual screen reader read-through because the automated scan came back clean.
• Treating Course Engagement findings as pass or fail the same way Accessibility findings are, when they're explicitly advisory.
• Assuming an AI-enhanced finding is infallible because it involved AI, rather than a fast first pass that still benefits from a human sanity check.

Practical checklist

• Confirm alt text is actually accurate to the image, not just present.
• Confirm learning objectives are clearly stated and genuinely met by the content.
• Judge whether an image or chart's meaning actually comes through, not just whether it has a description.
• Judge whether an interaction is genuinely purposeful, not just technically functional.
• Confirm tone, context, and language fit the intended audience.
• Run a real screen reader read-through before publishing, especially for a course used by a wide range of learners.

Frequently asked questions

Summary

A11yCheck's automated scan reliably catches structural, technical, and measurable accessibility and quality issues, fast and consistently. It can't judge instructional quality, learning objectives, or subjective meaning, that still takes a human. Knowing exactly where that line sits is what makes the tool worth trusting.