How Accessibility Gets Evaluated

Now that you know what accessibility means and why POUR matters, how do you actually tell whether a site meets the bar?

WCAG structure basics

The Web Content Accessibility Guidelines (WCAG) are organized into three conformance levels:

• Level A — the minimum. Without this, some users are completely blocked from using the content.
• Level AA — the standard most laws and organizations require, including AODA and most Canadian provincial legislation. This is the level most courses, audits, and policies target.
• Level AAA — the highest level. Comprehensive but often impractical to apply site-wide; typically targeted for specific content types rather than an entire site.

When someone says a site is “WCAG 2.2 AA compliant,” they mean it meets all Level A and Level AA success criteria in that version of the guidelines.

Automated vs. manual vs. user testing

This is one of the most misunderstood parts of accessibility work, so it’s worth being direct about it: automated tools only catch a fraction of accessibility issues — commonly cited estimates put it around 30-40%. They’re excellent at flagging things like missing alt text, insufficient color contrast, or missing form labels. They cannot tell you whether alt text actually makes sense, whether a keyboard trap exists, or whether your content is genuinely understandable.

That’s why real evaluation combines three approaches:

• Automated testing — fast, scalable, catches structural and technical issues
• Manual testing — a person checking keyboard navigation, focus order, and reviewing whether automated flags are accurate
• User testing — people who actually use assistive technology testing the real experience

Relying on just one of these gives you an incomplete — sometimes misleading — picture.

Video (90 sec)Split-screen or sequential clip: an automated scanner reports a page as “passing,” followed by a manual/keyboard walkthrough revealing a real barrier (e.g., a focus trap or illogical tab order) the scanner missed entirely.

Video summary (text alternative): [The scanner gave a clean report, but manual testing found a keyboard trap the automated tool couldn’t detect — illustrating why layered testing matters.]

Interactive: Automated results vs. manual resultsToggle between “automated scan results” and “manual test results” for the same mock page — the scan reports 0 issues while the manual test flags 2-3 real problems.