The product page now exists in several languages, and the question arises: how do a search engine and a person find the right version. The answer starts with addresses. Every indexable language version must have its own stable URL - one that can be opened, bookmarked, and returned in results. The hreflang markup is a hint to the search engine about which URL serves which language and region; it translates nothing on its own, it only links ready versions together.

The choice of address structure comes down to three strategies, and all further work depends on it.

A subdirectory like /ru/docs is usually the easiest to maintain: one domain, one certificate, shared infrastructure. A subdomain ru.example.com gives more isolation but needs separate setup. A dedicated country domain example.de is the strongest market signal and the most expensive to run. What matters is picking one scheme and sticking to it: a mixture confuses both crawlers and analytics.

A separate trap is the automatic redirect by the Accept-Language header. It is convenient, but it must not cut the user off from other versions. Googlebot crawls mostly from the US, and a hard redirect can hide the localized pages from it entirely. The right move is to treat the detected language as a hint and offer a switcher, rather than force the user away or block direct links to specific versions.

What links the versions to each other is hreflang. It works on reciprocity: each version links to itself and to all the others. If page A points to B but B does not point back to A, the cluster is considered incomplete and the engine may ignore the markup. Hence the practical rule: the set of hreflang links is identical on every page of the cluster and includes the current page itself.