Accessibility Testing: What Automated Scanners Miss

• —Images with no alternative text attribute at all.
• —Text and background colour combinations that fail contrast ratios.
• —Form fields with no programmatic label.
• —Common misuse of ARIA attributes and missing page language.
• —Duplicate identifiers and some heading-structure problems.

These are real issues, they are common, and they are cheap to find. Running a scanner in your build pipeline stops many of them from ever shipping.

Most of the issues that stop people from completing a task are about meaning and flow, and those need a person.

• —Whether alternative text is useful. A scanner sees that an image has alt text, not that the text says "image123.png" or describes nothing the user needs.
• —Whether focus order is logical. The order can be technically valid and still jump around the page in a way that confuses.
• —Keyboard traps and hidden focus. A custom widget may work with a mouse and strand a keyboard user.
• —What a screen reader actually announces. The experience of a dialog, a live update or a custom control can only be judged by using it.
• —Whether errors and instructions are clear. A message can exist and still be unhelpful.
• —Cognitive load. Dense layouts, ambiguous labels and time limits affect people in ways no scan measures.

Version 2.2 of the Web Content Accessibility Guidelines added nine success criteria and retired one (4.1.1 Parsing). Several of the new ones concern how people interact, which is exactly where scanners are weakest:

• —2.4.11 Focus Not Obscured (Minimum), AA: a focused element must not be entirely hidden, for example behind a sticky header.
• —2.4.12 Focus Not Obscured (Enhanced), AAA and 2.4.13 Focus Appearance, AAA: stronger requirements on focus visibility.
• —2.5.7 Dragging Movements, AA: anything done by dragging needs a simple alternative.
• —2.5.8 Target Size (Minimum), AA: pointer targets must be large enough or spaced adequately.
• —3.2.6 Consistent Help, A: help mechanisms appear in a consistent place across pages.
• —3.3.7 Redundant Entry, A: do not make people re-enter information they have already given.
• —3.3.8 Accessible Authentication (Minimum), AA and 3.3.9 Enhanced, AAA: sign-in must not depend on a cognitive test such as remembering or transcribing a password, with allowances.

Always check the W3C's own text for the exact wording and levels; this list is a summary.

1. Scan automatically on every build and on a full-site sweep, to catch the repeatable issues early.
2. Group the findings by component, so one underlying fix clears many instances.
3. Test by keyboard. Work through key journeys using only Tab, Shift+Tab, Enter, Space, Escape and the arrow keys. Can you reach everything? Can you always see where you are? Can you always get out?
4. Test with a screen reader on the journeys that matter most, on at least one desktop and one mobile combination you expect users to have.
5. Check zoom and reflow. Increase text size and browser zoom and look for clipped or overlapping content.
6. Judge the content. Are alternative texts meaningful, labels clear, errors helpful?
7. Report by criterion and retest. Show findings against the specific success criteria, and retest fixes rather than assuming them.