Findings
Severities, results, evidence, and how findings are ordered.
A finding is one check, with a result. Everything below the score is findings, and they are the part of the report worth acting on.
The three severities
| Severity | Means | Examples |
|---|---|---|
| Critical | Actively preventing the page from being indexed or ranked. Fix these before anything else. | A noindex on a page you want indexed, a missing or empty title, a redirect loop, an expired certificate. |
| Recommended | A real improvement. Not on fire, but worth doing, and this is where most of the score lives. | A description that will be truncated, a skipped heading level, images without dimensions, missing structured data. |
| Notice | Worth knowing, worth doing while you are already in the file. Does not move the score much. | Underscores in the URL, a legacy image format, a missing llms.txt. |
Severity is about the page, not about you. A critical finding is critical whether or not it is easy to fix, and a recommendation stays a recommendation even when it takes five minutes.
Pass, fail, skip
| Result | Means |
|---|---|
| Pass | We checked, and the page is correct. Listed so you can see the check ran. |
| Fail | We checked and found a problem. The severity says how much it matters. |
| Skipped | We could not check. Field vitals need real traffic before they exist, and headers cannot be read from a connection that failed. Never counted against the score. |
What each finding gives you
| Field | What it is |
|---|---|
| Label | What was checked, as a noun. "Canonical tag", "Title length". |
| Message | What was found, in a sentence. |
| Evidence | The actual value: the title we read, the URLs in the chain, the images concerned. |
| Remediation | What to change. Specific where the fix is specific. |
| Impact and effort | Two numbers from 1 to 5, described below. |
| Vector | A stable identifier such as title.pixel_width. It does not change between releases, so it is safe to key your own tooling on. Visible in the API response. |
Impact and effort
Most findings carry two numbers from 1 to 5. Impact is how much fixing it is likely to matter. Effort is how much work it is, estimated generously: a meta tag is a 1, restructuring how a page renders is a 5.
The report is ordered by both. A finding with impact 4 and effort 1 is placed above one with impact 5 and effort 5, because the first is the one you will finish this afternoon. If you would rather work purely by impact, the ordering is yours to ignore, and the API returns both numbers on every finding so you can sort them however you like.
Over the API every finding arrives with its vector, severity, result, impact, effort and evidence as separate fields, so you can filter to criticals, raise a ticket per finding, or fail a build on a specific vector. See Reports.
Evidence
Where a finding can point at something concrete, it does: the title it found and its measured width, the four images missing alt text and their filenames, the exact redirect chain it followed. A finding that tells you something is wrong without showing you what it saw is a finding you have to verify by hand, so we avoid producing them.