We re-examined the Page Structure Diagnostic for what it measures, which page it looks at, and whether applying its suggestions changes the diagnosis — and corrected five judgment criteria.
- List-item counting — lists in menus and footers outside the main content were being counted; counting is now limited to the main content.
- robots.txt allow rules — adding the allow rule the diagnostic suggested still read as "blocked" on re-check. Allow rules now take precedence per the standard, path rules are interpreted, and the report shows which rule caused a block.
- Page language detection — the display language used to override the page's language, so an English page viewed in Korean was scored against Korean thresholds. The language is now detected from the page itself, and you can set it explicitly in the input form.
- Page-type-aware schema recognition — product, service, about, home, and listing pages are recognized when they carry structured data that fits their type, even without an article-style schema.
- Product offer relationships — an offer attached to a recommended product could mask a missing offer on the main product. Only offers linked to the main product now count.
Because the criteria changed, scores and grades may go up or down, and comparisons against past results restart from your first diagnostic under the new criteria. Previous scores and "applied" marks stored in your browser are reset as well. Once you apply a suggested fix and re-run the diagnostic, that item is judged as resolved.