Skip to main content
This page lists every check in the localisation audit. The weight sets how much a check counts toward its dimension and the overall score. The How it is scored column uses three labels:
  • Measured: scored from page data the same way on every run.
  • Measured, AI fallback: measured when the evidence is clear. When it is ambiguous, an AI review is shown but not scored.
  • AI-assessed: reviewed by a language model and shown under AI-assessed, not scored. It never changes the score.
“Share of pages” means the score is the percentage of inspected pages that pass.

Technical

hreflang applies only when the audit finds more than one locale.

Linguistic

Contextual

Visual

The three browser checks only fail a localized page when both pages have the same viewport measurement and the source-language version is cleaner. A layout bug that also exists on the source page is a design issue, not a localisation issue, so it is not counted. Pages without a matching source viewport are skipped. Elements inside carousels and other horizontally scrolling containers are skipped, because overflow there is usually intentional. These checks are not applicable when layout measurements are not available for any localized page.