Transparent about what the scanner knows—and what it cannot know
The system records reproducible automated results and clearly separates them from checks that require human judgment.
A successful automated scan does not mean a page or website conforms to WCAG, EN 301 549, the EAA, BFSG, BITV, or any other legal requirement.
Pass or fail, no guesswork: alt text, contrast, heading order, form labels, ARIA misuse.
No strict verdict — a recommendation worth applying, such as link wording.
Captions, focus order, text inside images, whether content actually makes sense.
Scope and discovery
You tell us which website to check. We then follow its sitemap to find every public page — but only within limits that keep the scan safe and predictable. Any page we could not reach, or chose not to check, is listed, not hidden.
- Only websites you confirm you own or manage
- No scanning of private networks or internal systems
- Stays on the same website — it will not wander off to other domains
- Pages left out of the scan are always listed
Render and evaluate
A real browser opens each page exactly as a visitor's browser would, including anything only JavaScript builds. It then runs a fixed, versioned set of accessibility checks against what actually loaded — not against the raw source code. Every check falls into one of three groups, explained in the table below: automatically confirmed, best-practice guidance, or a note that a person should look.
Group repeated problems
The same problem often shows up on many pages — the same missing label on every product page, for example. Instead of listing it fifty times, we combine it into one item that keeps every affected page, the exact WCAG rule, a plain-language explanation, and a suggested fix, without ever deleting the page-specific detail underneath.
WCAG 2.2 combined
Score v6 is a risk-weighted assessment of automated WCAG 2.2 checks, not a percentage of criteria passed. A and AA are calculated separately and weighted equally; the overall score is the lower of this average and the pooled risk score. Each applicable rule per page provides four baseline points. Failures cost 1/3/7/12 points for minor/moderate/serious/critical problems, multiplied by min(4, 1 + log2(occurrences) / 2). The severity × repetition loss is capped at 8 units per rule per page before level weighting. In the pooled score, A baseline and loss carry a 1.5 weight; AA carries 1. Each component equals max(0, 100 × (1 − loss / baseline)). Unassessed levels receive no invented score. Occurrences come only from confirmed violations; manual-review items, best practices, and AAA remain outside the score. Rule-set version and page coverage are shown separately. Even 100 points do not establish complete WCAG conformance.
Verify
Once a fix is made, a targeted re-scan checks exactly that page and rule again. The original failure stays on record right next to the new result, so nobody has to just take someone's word that it is fixed.
- A fresh, dated result every time you verify
- Nothing gets silently overwritten or deleted
- A failed or incomplete scan is recorded too, not hidden
What's automated, and what needs a guiding hand
Most accessibility problems — around nine in ten, in our experience — are caught automatically, with no guesswork. A smaller set of checks carry a "best practice" label: there is no strict pass or fail, just a recommendation worth applying. A short list of checks genuinely need a person to look, because a machine cannot judge them reliably.
| Check | How it is checked |
|---|---|
| Image alt text present | Automated |
| Colour contrast of text | Automated |
| Heading order and structure | Automated |
| Form labels and error messages | Automated |
| Duplicate IDs and incorrect ARIA usage | Automated |
| Link and button wording clarity | Best practice |
| Consistent navigation and layout | Best practice |
| Contrast of text baked into images or video | Needs a person |
| Captions and audio descriptions for video | Needs a person |
| Keyboard focus order and traps | Needs a person |
| Whether complex content actually makes sense | Needs a person |
A machine cannot always tell whether instructions make sense, whether a video needs captions, or whether text baked into an image has enough contrast — exactly the checks marked "Needs a person" above. Everything else, the scanner catches on its own.
The standards behind these checks
OpenForAll's automated rules are built around WCAG 2.2, the technical standard referenced by the laws and regulations below. We link to the official source for each — automated scanning does not by itself establish compliance with any of them.
Use automation as one layer of evaluation.
Combine recurring scans with keyboard, screen-reader, content-quality, and user testing by qualified people.