International SEO: the right version for the right market.

Google separates two cases. A multilingual site offers content in more than one language; a multi-regional site explicitly targets users in different countries. Many sites are both (Managing Multi-Regional and Multilingual Sites). For all of them, Google recommends a separate URL per version and hreflang annotations, which can be given in HTML, in HTTP headers or in a sitemap. Google calls the three methods equivalent and says using more than one has no benefit in Search (Localized Versions of your Pages).

The rules are specific. Each version must list itself and every other version, using fully-qualified URLs. Annotations must be reciprocal: If two pages don't both point to each other, the tags will be ignored. Codes are a language in ISO 639-1 format with an optional region in ISO 3166-1 Alpha 2 format, and a region alone is not valid. Reserved or wrong region codes, such as UK, have no effect. Where several URLs target one language in different locales, Google suggests a catch-all page for that language, and recommends x-default as the fallback for unmatched languages, designed for language selector pages. Google does not use hreflang or the HTML lang attribute to detect a page’s language.

We cover hreflang and market structure as part of technical SEO. On the technical SEO page, hreflang sits under the question of which version counts, alongside canonicals, duplicates and parameters.

Google compares four URL structures. A country-specific domain such as example.de gives clear geotargeting and makes server location irrelevant, but it can be expensive, may carry registration restrictions, and targets only a single country. Subdomains on a generic domain are easy to set up and separate, but users may not recognise the geotargeting. Subdirectories such as example.com/de/ are easy to set up and low maintenance on the same host, but separation of sites is harder and there is a single server location. URL parameters such as ?loc=de are not recommended (Managing Multi-Regional and Multilingual Sites).

Google lists the signals it uses to find a page’s target audience: country-code top-level domains, hreflang, server location and other signals such as local addresses, phone numbers, language, currency and links from local sites. Some country-code domains, such as .tv and .me, are treated as generic, as are .eu and .asia, so these need an explicit target. Language subdomains such as en-gb.example.com are not used to determine the audience, so the mapping must be stated explicitly. Google ignores locational meta tags such as geo.position. Crawling is another consideration, because each hostname has its own crawl budget (Crawl Budget Management).

We cover domain and folder strategy as part of technical SEO, and it is one of the decisions our international launch projects deal with, alongside hreflang and localisation. See SEO projects.

Google says localised versions of a page are only considered duplicates if the main content remains untranslated, so a fully translated page is not a duplicate of its source (Localized Versions of your Pages). The risk sits in partial localisation. If only the template, such as navigation and footer, is translated while the body stays in one language, the same content can appear multiple times in results with different boilerplate. Google recommends a single language for content and navigation on each page and avoiding side-by-side translations.

Same-language regional variants, for instance US and UK English, are duplicates in Google’s terms. Google advises using both canonicalisation and hreflang so that the correct regional URL is served, with the canonical in the same language (What is URL Canonicalization). Redirecting users automatically by language can stop users and search engines reaching all versions. Googlebot usually crawls from the USA without an Accept-Language header, and Google advises against IP analysis for adapting content (Managing Multi-Regional and Multilingual Sites).

Automation has a limit. Google’s spam policies list creating many pages through automated transformations, such as translating, where little value is provided to users, as scaled content abuse (Spam Policies for Google Web Search). We cover localisation without duplication as part of technical SEO, as one of the named parts of our international launch projects.

Google’s guidance on site moves applies strongly to international structures, because a move that changes domains or paths touches every market at once. Its advice includes preparing a mapping from current to new URLs, testing the new site thoroughly, redirecting old URLs to new ones with server-side permanent redirects, and monitoring traffic on both. Where a site is large, Google suggests moving a piece first, and it advises changing only one thing at a time (Site Moves and Migrations).

International moves add dependencies. Google says that if moved pages use hreflang annotations, they must be updated to the new URLs, and that each new URL should have a self-referencing canonical. Because hreflang depends on reciprocal links, a half-updated cluster can leave annotations ignored. Google also expects temporary ranking fluctuation, with medium-sized sites taking a few weeks or more for new URLs to be shown, and longer for larger ones, and it points to Search Console, where each property is verified separately.

We cover multi-market migrations as part of technical SEO. Our technical SEO FAQ says we handle migrations: planning, redirect mapping, testing, launch monitoring and recovery, for single sites and multi-market structures. The migrations guide covers the general sequence.