- Testing
- WCAG
- Manual review
What automated accessibility testing can't catch
Automated tools find roughly 57% of accessibility issues, according to Deque's study. Here is what they find well, what still needs people, and how to plan a manual review.
The Allyfolio team
· 2 min read
Automated accessibility testing is fast, cheap and repeatable – and it is only part of the picture. Deque analysed nearly 300,000 issues from more than 13,000 pages in first-time audits and found that 57.38% of issues by volume were identified by automated tests using the open-source axe-core rules (Deque: The Automated Accessibility Coverage Report). The rest needed people.
This article is general information, not legal advice.
What automated checks do well
- Colour contrast of text against its background (WCAG 1.4.3) – the single biggest category in Deque’s data.
- Missing names for buttons, links and form fields (4.1.2, 3.3.2).
- Missing alt attributes on images (1.1.1).
- Missing page language (3.1.1) and missing page titles.
- Structural problems such as skipped heading levels or invalid ARIA.
These issues are common, so catching them automatically on every release removes a lot of work.
What still needs a person
- Meaningful alt text. A tool can see that
altexists. It cannot tell whether “image123” describes a blue running shoe. - Focus order and focus visibility (2.4.3, 2.4.7). In Deque’s data, every focus order and focus visible issue was found manually.
- Keyboard operation (2.1.1) of custom menus, sliders and date pickers.
- Understandable content – clear instructions, helpful error messages, consistent navigation.
- Captions and audio descriptions for video – whether they exist and are accurate.
- Whole journeys. Can a screen-reader user actually find a product, pay and receive confirmation?
- New WCAG 2.2 criteria such as Focus Not Obscured, Dragging Movements and Accessible Authentication often need human judgement – see EN 301 549 and WCAG 2.2.
Plan a manual review
- Run automated scans on your key templates and journeys (start with a free scan).
- Fix what they find – it is the quickest win.
- Walk through your main journeys with only a keyboard, then with a screen reader.
- Review alt text, link text, headings and error messages by reading them.
- Write down what you tested, when and what you changed. A dated record is useful evidence of your work under the EAA and BFSG.
A clean automated scan is a good start, not a finish line – and no overlay widget can do the manual part for you. Allyfolio’s plans combine scheduled scans with a checklist for the manual review that automation cannot do.