Normes référencées
Besoin d'un audit technique rigoureux de vos produits numériques au regard de ces normes ?
Demander un auditThe 100% Automated Compliance Myth
Vendors selling automated widgets and scanning bots frequently promise '100% automated compliance'. In reality, automated rules engines (including axe-core, Lighthouse, and commercial crawlers) can only evaluate rules that can be determined algorithmically without human context. Extensive industry studies demonstrate that automated tools reliably identify between 30% and 40% of WCAG defects.
What Automated Scanners Do Well
Automated tools excel at catching objective, rule-based syntax errors at immense scale across thousands of pages:
- Color contrast calculations on static background/foreground text pairings.
- Missing alt attributes on img elements.
- Duplicate HTML id attributes in the DOM.
- Form controls with no associated label element or ARIA name.
- Missing lang attribute on the root html element.
The 60–70% Scanners Completely Miss
Automated scanners cannot evaluate human comprehension, dynamic context, or interactive state transitions:
- Quality of alternative text: A scanner confirms alt='dog' exists, but cannot know if the photo is actually a corporate org chart.
- Logical keyboard focus order: A scanner cannot evaluate if tabbing jumps chaotically across the screen.
- Keyboard traps: A scanner cannot detect if a user gets stuck inside an open modal dialog.
- Error message clarity: A scanner cannot determine if an error message clearly explains how to fix the mistake.
- Screen reader announcement accuracy: A scanner cannot verify whether dynamic live regions speak the right message at the right time.
- Responsive reflow usability: A scanner cannot determine if content overlaps or disappears at 400% zoom.
The Ideal Hybrid Engineering Framework
The most effective enterprise approach uses automated tests in CI/CD to prevent regressions of objective rules, while deploying experienced accessibility engineers to perform manual keyboard, screen reader, and edge-case testing before major releases.
TaggedWorks Research
Pôle de recherche et d'ingénierie en accessibilité numérique
Les publications de recherche de TaggedWorks sont rédigées par des ingénieurs et spécialistes praticiens. Nos préconisations reflètent les interprétations actuelles des WCAG 2.2, les comportements des navigateurs et les matrices de compatibilité des technologies d'assistance.