What automated accessibility testing can and cannot tell you

Automation finds evidence, not a legal guarantee

Automated testing can identify many markup, name, relationship, contrast, and ARIA problems quickly. It cannot decide every question of meaning or usability and should not be marketed as instant compliance.

False positives and false negatives both matter

A trustworthy scanner must verify that the real page loaded before running rules, show its coverage, and stop when a bot challenge or empty browser document makes the result unreliable.

It must also be honest about surfaces it did not reach, including gated checkout steps or interaction states.

Use three layers of evidence

Combine deterministic automated checks, guided manual review, and periodic testing by people using keyboards and assistive technology. Each layer answers questions the others cannot.

• Automated: repeatable rule checks and regression detection.
• Guided review: intent, labels, reading order, and interaction judgement.
• User evaluation: real task completion and lived experience.