What an Accessibility Audit Actually Checks (And What It Can't)
"Accessibility audit" gets thrown around a lot — by plugin vendors, consultants, and lawsuit demand letters. But what does an automated audit actually examine, and where does it stop? Here's the honest breakdown.
What automated checks catch reliably
Missing text alternatives. Every image needs either meaningful alt text or an explicit empty alt if it's decorative. Scanners catch missing alt attributes with near-perfect accuracy.
Unnamed controls. Buttons and links that contain only an icon — no text — are invisible to screen readers unless they have an accessible name (aria-label or equivalent). Scanners find these reliably.
Unlabeled form fields. Every input needs an associated label. Scanners check the label-input association programmatically.
Missing page language. The <html> element should declare the page's language so screen readers use the right pronunciation. Trivial to check, commonly missing.
Structural basics. Skipped heading levels, duplicate IDs, missing document landmarks — all mechanically detectable.
Keyboard traps (some). Scanners can detect some focus-management failures, though not all.
What automated checks can't do
Color contrast — partially. Tools can compute contrast ratios, but they can't tell you whether the combination is actually readable in context, or whether information is conveyed by color alone.
Meaningful alt text. A scanner can tell alt text is missing. It cannot tell you whether existing alt text is accurate — "image123.jpg" passes the mechanical check and fails the human one.
Heading logic. A scanner sees you skipped from h2 to h4. It can't tell you whether your heading structure makes sense as an outline of the page.
Complex interactions. Carousels, modals, multi-step checkout flows, drag-and-drop — these need a human (ideally one who uses assistive technology) to evaluate.
The actual user experience. No scanner tells you what it's like to buy something on your site with a screen reader. That's the test that matters, and it's irreducibly human.
How to read an audit report
When you get audit results — from our free scan or anyone else's — read them in three buckets:
- Auto-fixable: mechanical issues with safe, deterministic fixes. Fix them all; there's no reason not to.
- Needs human review: real issues where the fix requires judgment (what should this alt text say? does this heading order make sense?). Budget time for these.
- Can't auto-check: the experiential stuff. If accessibility matters to your business, get a human audit for these at least once.
Any vendor whose report has only bucket 1 — or who claims buckets 2 and 3 don't exist — is selling you a comforting fiction.
EqualFix's free plugin runs all of these checks against your WordPress site and sorts every finding into exactly these buckets. Try it.