Accessibility scans detect approximately 25% of WCAG issues. The remaining 75% requires human evaluation because those success criteria depend on context, meaning, and user experience that code analysis cannot interpret. This 25% figure holds across traditional scanning tools that evaluate HTML, CSS, and ARIA against WCAG success criteria.
| Key Point | What It Means |
|---|---|
| Scan detection rate | Approximately 25% of WCAG issues are detectable by automated scans. |
| Scan coverage | Scans evaluate programmatic attributes in HTML, CSS, and ARIA against machine-checkable rules. |
| What scans miss | Roughly 75% of WCAG issues require human judgment about meaning, context, and user flow. |
| Correct use of scans | Scans serve as a first pass and an ongoing monitoring layer, not a conformance verdict. |
Where the 25% Figure Comes From
WCAG success criteria fall into two broad categories: those that can be evaluated programmatically and those that require human interpretation. A scan loads a page, parses the HTML, reads CSS and ARIA, and checks attributes against rule sets that map to WCAG. That process works well for criteria with clear pass or fail conditions expressed in code.
The 25% figure reflects the portion of WCAG success criteria that can be reliably evaluated this way. It is a consistent benchmark used across the accessibility field when describing what scans can and cannot do.
What Scans Can Detect
Scans are accurate and useful within their coverage area. They identify issues where the presence, absence, or value of a code attribute determines conformance.
- Missing alternative text on images where the alt attribute is absent
- Empty form labels or inputs with no associated label element
- Missing document language declarations in the HTML root
- Duplicate ID values in the document
- Missing page titles or empty title elements
- Invalid ARIA usage such as roles paired with unsupported attributes
Within this set, scans produce high-confidence results. They run fast, return consistent output, and scale across large sites through scheduled monitoring.
What Scans Cannot Detect
The remaining 75% of WCAG issues require human evaluation because the criteria depend on meaning. A scan can confirm that an image has alt text, but it cannot confirm the alt text accurately describes the image. A scan can confirm a heading exists, but it cannot confirm the heading structure reflects the actual information hierarchy of the page.
Criteria involving keyboard flow through an interactive component, the clarity of error messages, the logical order of content, the accuracy of link text, and the usability of a custom widget all fall outside what code analysis can verify. Evaluating these criteria requires a person using assistive technology, inspecting the interface, and exercising judgment about the user experience.
How to Use Scans Correctly
Scans work best as the first layer in a broader evaluation strategy. They flag the machine-checkable issues quickly, which gives development teams a running feed of problems to remediate between formal audits. Scheduled scans monitor pages on a recurring cadence to catch regressions introduced by new deployments or content updates.
A scan result of zero issues does not indicate WCAG conformance. It indicates the scan identified no issues within its 25% coverage area. Full conformance verification requires an audit that combines scanning with screen reader testing, keyboard testing, visual inspection, and code inspection conducted by accessibility professionals.
Why the Percentage Matters for Planning
Teams that treat a clean scan as proof of accessibility build compliance programs on a foundation that covers one quarter of the actual requirements. Budget, timeline, and staffing decisions should reflect the full picture: automated scans for continuous coverage of the machine-checkable issues, paired with periodic audits that evaluate the criteria scans cannot reach.
The 25% figure is not a limitation to work around. It is a statement of what a specific tool category does well, alongside what other evaluation methods are designed to cover.