Errors
An error is a confirmed failure of a WCAG success criterion: an image with no description, text that fails contrast, a button with no name. Errors lower the index.
Each error has a severity, based on how badly it blocks people:
| Severity | What it means | Example |
|---|---|---|
| Critical | Some people can’t complete a task at all | A form you can’t submit with a keyboard |
| Serious | A task is very hard, or information is lost | Links a screen reader announces only as “link” |
| Moderate | A task is harder than it should be | Low contrast on secondary text |
| Minor | An annoyance or inconsistency | A redundant description repeated beside a photo |
Alerts
An alert is something a machine can flag but only a person can judge. For example: “this text sits over a photo; check its contrast”, or “these two images share the same description; is that right?”
Every alert gets a decision, recorded with who made it, when and why:
- Fixed: we changed the site; the next scan confirms it.
- Verified: a person checked it and it’s fine as it is. For example, text over a photo measured against the darkest and lightest pixels behind it.
- Accepted risk: a real limitation we can’t fix right now, such as a third-party widget. Your accessibility statement says so.
- Won’t fix: out of scope, with a reason.
- False positive: the tool was wrong about this element.
Decided alerts leave the “open” count but stay listed in the report under Checked by hand, so nothing is hidden. If the page behind a decision changes, the decision is flagged for re-checking.
Grouping
A raw scan reports every instance: the same header button on 30 pages is 30 instances. Reports group instances by root cause, so that becomes one issue with 30 pages listed, and one fix.
Confidence
- High: two independent engines flagged the same element as a failure.
- Medium: one engine flagged it.
- Review: an engine says it might be a problem. It becomes an alert for a person to decide.