Accessibility Testing, Explained

Automated scanner, manual audit, or real screen reader? Here's what each one actually catches.

July 9, 2026 · 6 min read

Someone asked us what WAVE is — a well-known free accessibility scanner — and whether it's enough to know a site works for a blind visitor. Short answer: it's a real, respected tool, and it still can't tell you that. Here's the honest breakdown of what each level of testing catches, with real examples from our own audits.

Three tiers, three different ceilings

Every accessibility check falls into one of three tiers. Each one catches real problems the tier below it misses — and each one has a ceiling it can't see past, no matter how thorough it is within its own tier.

1

Automated scanners

Examples: WAVE, axe, Google Lighthouse / PageSpeed Insights' Accessibility score, our own free ADA Compliance Checker

These read the page's code and flag patterns that are reliably detectable by a machine — missing alt attributes, missing form labels, insufficient color contrast, invalid ARIA, skipped heading levels. Fast, free or cheap, and genuinely useful as a first pass.

Catches
  • Missing alt attributes
  • Color contrast ratios
  • Missing form labels
  • Invalid or misused ARIA
  • Heading structure gaps
Misses
  • Alt text that's present but meaningless
  • Keyboard target size
  • Motion with no pause control
  • Focus hidden behind sticky elements
  • Whether it actually makes sense out loud
2

Manual WCAG audit

A person checks the page against the actual WCAG 2.2 success criteria by hand — not just the automated score.

This catches everything a scanner can, plus the categories of issue that only show up when someone actually interacts with the page: tabbing through it, measuring rendered target sizes, checking whether motion can be stopped, checking whether a sticky header ever hides a focused element. It still can't tell you what a screen reader announces, because a sighted person reading code isn't hearing what assistive technology actually outputs.

Real example: our own homepage scored 100/100/100/100 on PageSpeed's automated accessibility check. A manual pass still found three real issues underneath that score — undersized nav targets, a hero video with no pause control, and a sticky header that could obscure a focused section. Full before-and-after here.
3

Real assistive-technology testing

Testing with an actual screen reader (VoiceOver, NVDA, JAWS), keyboard-only navigation, or a blind or low-vision user.

This is the only tier that answers the question that actually matters: does this work for the person using it? Alt text can pass every automated check and still be useless. A form can have a technically valid label and still confuse someone navigating by ear. This tier is slower and harder to scale, which is exactly why it's the one most sites skip.

Real example: a business attorney's homepage had three images — including his own headshot — carrying unedited AI-generated placeholder alt text, literally reading "Here's an alt tag for the image: Headshot of a man in a suit." Every automated scanner would have marked that alt attribute as present and passing. A screen reader user would have heard the AI's instructions to itself. See the audit.
Another one: a live search box updated its results with zero feedback for anyone who couldn't see it change — on a page with a perfect automated accessibility score. Caught by hand, not by a scanner.
One more: a San Jose law firm's "Switch to ADA Accessible Theme" toggle loads a separate version of their site — and that version renders a literal, unremoved developer placeholder ("CHANGEME") right above the headline. No contrast ratio or missing alt attribute would ever flag that; it only surfaces when a person actually clicks the toggle and looks. See the audit.

Side by side

IssueAutomated scannerManual auditReal AT testing
Missing alt textYesYesYes
Color contrast ratioYesYesYes
Alt text present but meaninglessNoSometimesYes
Keyboard target sizeNoYesYes
Motion with no pause controlNoYesYes
Focus hidden behind sticky headerNoYesYes
Silent live-region updatesNoSometimesYes
Does it make sense read aloud, in orderNoNoYes

Which one do you actually need?

If you've never run any check at all, start with a free automated scan — it'll catch real, fixable issues in under a minute. If you're worried about ADA legal exposure or you've already fixed what a scanner found, a manual audit is the next real step, because that's the layer most demand letters and lawsuits actually target. Full assistive-technology testing with real users is the gold standard, and worth it for anything handling healthcare, government, finance, or legal services — the places where getting it wrong carries the most weight.

Start free

Run our automated checker on your own site right now — instant, runs in your browser, nothing sent to a server.

Check My Website →

Get the manual audit

A real, hand-reviewed WCAG 2.2 AA audit — the layer that catches what scanners can't. Starting at $499.

See the Audit →