Cross-Language Duplicate Content

Cross-language duplicate content is the international edition of SEO's most misunderstood topic, and it splits into two questions with opposite answers: do translations count as duplicates? (no — translated content is different content; relax) and do same-language market versions count? (functionally yes — your en-US and en-GB near-twins will be folded, filtered or mis-chosen unless the machinery says otherwise — the case this guide mostly exists for). Here's both halves, plus the third case (one language reused across structures) that catches ccTLD portfolios.

Case one: different languages — the non-problem

A German translation of your English page is, to the systems, a different document — translation is transformation, and no duplication treatment applies (the general rules' sameness test fails immediately across languages). What the pair still wants is the sibling relationship: hreflang linking them serves the right version per searcher and pools the cluster's understanding — wiring for service, not defence against a penalty that was never coming. The one caveat lives at the quality layer, not the duplication layer: raw machine translation at scale is a thin-content problem wearing a language costume.

Case two: same language, multiple markets — the real one

en-US/en-GB/en-AU versions differing by spelling and currency are near-duplicates by any textual measure, and untreated they behave like the standard case: Google picks one representative (not necessarily per-market), the others fold, and markets get mis-served — the flip-flop and wrong-version symptoms the error catalogue logs. The treatment stack: hreflang as the primary tool (this exact case is what regional codes exist for — declaring the twins as market siblings, each self-canonical, per the two-systems-agreeing rule); genuine differentiation as the reinforcement (the more truly local each version is — prices, law, examples, vocabulary — the more the systems and the buyers both treat them as distinct, which is the create-versions-only-where-differences-are-real doctrine from the planning rules); and restraint as the prevention — markets that don't justify differentiated versions get served by one geo-broad version instead of cloned folders manufacturing the problem.

Case three: one language, several properties

The ccTLD portfolio's trap: the same English content republished across example.com, example.co.uk and example.com.au — cross-domain near-duplication where the standard folding picks winners across your own properties (usually the strongest domain, starving the others). Same treatment logic — hreflang works cross-domain (full URLs, the reciprocity law across properties), differentiation per market, and the structural question honestly revisited: portfolios maintaining triplicate content without per-market operations are the consolidation-candidate profile the structure guide names.

Frequently asked questions

Will Google penalise our site for having the same product description in six English markets?

Penalise, no — the no-penalty rule holds internationally; the cost is selection and dilution (one version chosen, markets mis-served, signals split), which hreflang-plus-localisation converts into deliberate service. The fear is misplaced; the wiring isn't optional.

Should we rewrite content differently per English market to be safe?

Rewrite for localisation (vocabulary, examples, offers — the differences that serve buyers), never for artificial uniqueness — spinning paraphrases across markets is effort spent making every version worse, and the systems fold rewritten near-duplicates anyway. Real differences or shared text plus wiring: both honest; the fake middle serves no one.

Does translated content need canonical tags pointing at the original?

No — that's the case-one confusion in canonical form: cross-language canonicals tell Google to index only the original and drop the translation (occasionally what a token-translation page wants; never what a market version wants). Translations self-canonical, hreflang across — the standing pattern, after which each version competes in its market on the pair this silo always lands on: local content quality and local authority (our contribution).