Every audit issue carries a severity. We use three levels:
- Info — informational. Something to know, not something to fix urgently. Example: a non-canonical alternate href that’s harmless.
- Warning — suboptimal but not broken. Example: missing meta description, a thin canonical, an H1 longer than ~70 chars. Fix when convenient.
- Error — actively hurting. Example: 5xx response, missing canonical on duplicate-content paths, broken JSON-LD syntax. Fix soon.
What the Schema tab does with severity
Issues are grouped severity-first (errors at the top). Each row has a one-line message, a fix hint, and — if applicable — an Ask agent to fix button that drafts a JSON-LD replacement or a meta-tag change for you (1 credit).
How findings reference severity
The three Key Findings on a Run Detail page weight error-severity issues highest when synthesising recommended actions. A run with no errors but a dozen warnings will produce findings about citation gaps; a run with active errors will produce a finding calling them out specifically.