What can automated accessibility scanners detect?

Automated accessibility scanners are useful and genuinely limited. The limitation is structural. It is not a matter of the tool needing to improve. A large share of WCAG’s success criteria require human judgement about meaning and context. Software cannot evaluate that judgement, no matter how sophisticated the tool.

What automated accessibility testing can reliably check

Category	Example	Why software can check it
Presence of required markup	Does an image have an alt attribute at all?	A structural, yes/no check on the underlying code
Measurable values	Does text meet the minimum contrast ratio?	A calculation against known colour values
Structural patterns	Are heading levels used in a valid sequence?	A pattern that can be parsed from the document structure
Some ARIA misuse	Are accessibility attributes used in technically invalid ways?	Checkable against the technical specification

What automated testing tools cannot reliably check without manual accessibility testing

Category	Example	Why it requires human judgement
Meaning and accuracy of alt text	Does the alt text actually describe what the image conveys?	Requires understanding the image’s content and its purpose on the page
Logical reading order	Does the heading structure make sense as an outline, not just follow valid syntax?	Requires understanding what the content means, not just how it’s tagged
Genuine keyboard operability	Can every custom interactive component actually be operated without a mouse?	Requires actually attempting the interaction, not just inspecting code
Whether an error message is actually helpful	Is the error text specific enough to fix the problem?	Requires reading and judging the message’s clarity
Context-dependent contrast exceptions	Is a low-contrast element genuinely decorative and exempt?	Requires judging the element’s actual function on the page

Why the commonly cited estimate matters

The one-third-to-a-half estimate, and why the exact number matters less than the reason

Estimates of how much of WCAG automated tooling can reliably detect commonly land somewhere between roughly a third and a half of all success criteria. The precise figure matters less than the reason behind it. Many criteria are fundamentally about whether something makes sense to a person, not whether it follows a detectable pattern. This is not a shortcoming a better tool will eventually fix. It is a structural limit of what software can evaluate about meaning, judgement and human experience.

Why a “green” or “passed” scanner result is not a finding of conformance

A scanner reporting no errors has checked the portion of WCAG it is capable of checking, and found no issues within that portion. It has not evaluated the remaining criteria at all. A genuinely inaccessible site can return a clean automated result if its failures happen to sit entirely within the categories software cannot assess. Treating a clean scan as proof of accessibility, or worse, marketing it as “certified compliant”, misrepresents what the tool has actually done.

Why this site’s own accessibility tool is a questionnaire, not a scanner

This site’s accessibility self-assessment is deliberately built as a graded questionnaire answered by a person, not an automated scan producing a pass or fail. A pass/fail result a business might rely on before assessing genuine accessibility risk is itself a form of the misrepresentation described above. The tool grades what’s reported. It does not, and cannot, certify conformance.

What a genuine accessibility audit and web content accessibility tests add that automated checks cannot

Combining automated scanning with manual testing by a person

A proper accessibility audit combines automated scanning for what it’s reliably good at — contrast ratios, missing alt attributes, structural markup issues — with manual testing by a person. That means attempting keyboard-only navigation, testing with an actual screen reader, and judging whether alt text, error messages and heading structure genuinely make sense rather than merely existing. What that combined process actually involves is covered on commissioning an accessibility audit.

What to ask a supplier who offers an “automated accessibility check”: screen readers, keyboard and assistive technology

Ask specifically what proportion of WCAG success criteria the check actually covers. Ask whether any manual testing is included. A supplier who represents an automated scan alone as a compliance check, or uses words like “certified” or “guaranteed compliant” based on a scan result, is making a claim the underlying technology cannot support.

Why false negatives are more dangerous than false positives here

An automated scanner occasionally flags something as an issue that isn’t genuinely a problem in context. That is a false positive. It is merely a minor inconvenience to review and dismiss. Far more consequential is a false negative: the scanner reports no issue in an area it simply cannot evaluate. That can be misread as a positive finding, rather than an absence of finding at all. This asymmetry is part of why a clean scan result carries less weight than it appears to on first reading.

What third-party overlay products claim, and why to be sceptical

A category of third-party “accessibility overlay” products market themselves as adding compliance to an existing site through a single script. Some claim to detect and fix issues automatically, or even to provide legal protection. Given the structural limits described above — that a large share of WCAG criteria require genuine human judgement about meaning and context — no automated script can reliably achieve full remediation this way. Claims of automatic, complete compliance from such a product should be treated with the same scepticism applied to any other unverifiable marketing claim.

Why this site declines to make the claim it’s warning against

This site could describe its own accessibility self-assessment tool as a “compliance checker” to sound more authoritative. It deliberately doesn’t. Doing so would be exactly the kind of overstatement this page describes. The tool is named and framed as a questionnaire specifically to avoid implying a certainty the underlying method cannot support. That is a discipline applied consistently across this site’s claims about itself, not just its claims about other tools.

What to do next

Treat any automated scanner result, including a clean one, as a partial finding rather than a conformance verdict. Pair it with manual testing wherever a genuine accessibility position matters to your business. What that fuller process looks like is set out on commissioning an accessibility audit.