Are the language versions real translations

Reads each language version of the client's pages (the groups the crawl's hreflang or /es/-style paths tie together) and scores every pair against its source 0–100: equal H2 and H3 counts; title, meta description and schema inLanguage actually localised (the language detected, not just declared); length within the band the language should take (German 25–35% longer than English, Japanese 10–25% shorter, in words); a stale translation (modified before its source); numbers, dates, prices and phone numbers written for the locale; and source-language sentences left untranslated. Unmeasured parts are null and the rest rescale. Each pair also carries a cultural adaptation check against its market's profile (DACH, Francophone, Hispanic, Japanese: formal address, hard-sell CTAs, Impressum or Mentions légales, local number, date and currency formats, US brands, statistics and laws, length against English), each failure a Medium 'Cultural Adaptation' finding, and machine-translation QA (body identical to the source, <html lang> against the detected language, a translated meta description over 160 characters, schema inLanguage, Google's translate widget). contentParity scores each group 30/30/25/15: the page exists in every declared language, title, meta and schema localised, structure (sections ±1, images and charts exact, FAQ items ±2), and freshness (≤7, 8–30, 31–90, 90+ days behind). A heuristic, and it says so. url picks the group containing that page; sourceLang which version the others are measured against (default the x-default one). Answers 'is our German site actually translated' and 'which translations are out of date'. Free: fetches the client's own pages.