To rank a site in several languages, give every language version its own crawlable URL, connect the versions with hreflang annotations that link back to each other, point each page's canonical at itself in its own language, and list every variant in your XML sitemaps. Google's Search Central documentation, updated in September 2026, describes this as telling Google about localized versions of a page so Search can show each searcher the most appropriate language or region. The rest of this guide covers the decisions behind that setup and the mistakes that quietly break it.
If you are new to the basics, start with why SEO drives organic traffic; this post assumes you already have one language working and want to add more.
Multilingual vs multi-regional: know which one you are building
Google separates two cases. A multilingual site offers content in more than one language, such as a Canadian business with English and French pages. A multi-regional site explicitly targets users in different countries, such as a manufacturer that ships to both Canada and the United States. Many sites are both. The distinction matters because it decides whether your hreflang values need a region code (en-US, en-GB) or only a language code (en, de).
Choosing a URL structure
Google recommends separate URLs for each language version rather than switching content with cookies or browser settings. Its crawler usually originates from the USA and sends requests without an Accept-Language header, so content that only appears after detection may never be crawled. That leaves four structures, summarized from Google's own comparison:
Structure | Example | Strengths | Trade-offs |
|---|---|---|---|
Country-code domain (ccTLD) | example.de | Clear geotargeting; easy separation; server location irrelevant | Expensive, more infrastructure, sometimes strict registration rules; targets one country only |
Subdomain on a generic domain | de.example.com | Easy to set up; allows different server locations | Users may not know if "de" means language or country |
Subdirectory on a generic domain | example.com/de/ | Easy to set up; low maintenance on one host | Single server location; harder to separate sites |
URL parameter | example.com?loc=de | None worth relying on | Not recommended by Google |
For most businesses expanding from one language, subdirectories are the pragmatic default: one domain, one set of authority signals, one deployment. ccTLDs make sense when you run genuinely separate country businesses. Two details are easy to miss. Google treats some country-code domains, including .io, .co, .ai, .me and .tv, as generic, and it treats .eu and .asia as generic too, so they do not geotarget. And the language label in a subdomain or folder name is not used by Google to decide the audience; hreflang does that job.
hreflang rules that actually matter
You can declare alternates in three equivalent ways: link elements in the HTML head, an HTTP Link header (useful for PDFs), or xhtml:link entries in an XML sitemap. Google says using more than one method brings no benefit in Search and makes maintenance harder, so pick one.
1. Every version lists itself and all the others
Each language version must carry the complete set of alternates, including a self-reference. In practice the block of hreflang tags is identical on every version of the page, which makes it easy to generate from one source of truth.
2. Return links are mandatory
If page X names page Y as an alternate, Y must name X. Google ignores annotations that are not reciprocal, so a third-party site cannot claim to be a version of your page. Missing return links are the first item on Google's list of common hreflang mistakes.
3. Language first, region optional
The value is an ISO 639-1 language code, optionally followed by an ISO 3166-1 Alpha 2 region code: de, en-GB, fr-BE. A region on its own is invalid, because Google does not derive the language from a country. The classic trap is writing "be" for Belgium, which is actually the code for Belarusian. Codes outside those standards, such as es-419 for Latin American Spanish, are not supported, and reserved codes such as EU, UN or UK have no effect (the United Kingdom is GB). Scripts can be specified with ISO 15924, for example zh-Hans or zh-Hant.
4. Use fully qualified URLs
Write https://example.com/de/page, not /de/page or //example.com/de/page. Alternates may sit on different domains, which is how a ccTLD setup is connected.
5. Add x-default and a generic fallback
The reserved value x-default tells Google where to send users whose language matches none of your versions. Google designed it for language-selector pages, and it can point to one of your existing pages if you have no selector. If you have several regional variants of one language (en-IE, en-CA, en-AU), also provide a generic en page for everyone else.
A worked example
A store has English for the US, English for the UK, German, and a language selector at the root. Every one of the four pages carries the same five annotations:
hreflang="en-US" pointing to https://example.com/en-us/boots
hreflang="en-GB" pointing to https://example.com/en-gb/boots
hreflang="en" pointing to https://example.com/en-us/boots (the generic English fallback can reuse a specific page)
hreflang="de" pointing to https://example.com/de/boots
hreflang="x-default" pointing to https://example.com/
How canonical tags interact with hreflang
Canonical and hreflang answer different questions: canonical says which URL represents a set of duplicates, hreflang says which language or region version to show. They conflict when a translated page canonicalizes to the original language, because that tells Google the German page is a duplicate of the English one and removes it from consideration.
The rule from Google's canonicalization guide is to specify a canonical page in the same language when you use hreflang, or the best possible substitute language if no same-language canonical exists. In the normal case each language version is self-canonical. Where you do have true duplicates in one language, for example the same German content on example.de and example.com/de/, pick a preferred version and use rel="canonical" together with hreflang. Google also notes that, for canonicalization, it prefers URLs that are part of hreflang clusters.
Translated content vs duplicate content
Google states that localized versions of a page are only considered duplicates if the main content remains untranslated. A genuinely translated product page is not a duplicate of the original. Regional variants in the same language, such as en-US and en-GB pages that differ mainly in currency and spelling, are similar by nature; hreflang is how you tell Google they are intentional alternates.
What causes trouble is translating only the template. Google warns that translating the navigation and footer while keeping the main content in one language can create a poor experience when the same content appears several times in results with different boilerplate. Google determines a page's language from its visible content, not from the HTML lang attribute or the URL, so keep content and navigation in a single language per page and avoid side-by-side translations.
Per-language sitemaps
Sitemaps help discovery, and with the xhtml namespace they can carry the full hreflang set. Each url entry needs a loc plus xhtml:link children for every variant, including itself, so three versions mean three entries with three identical child lists. The child links do not count towards the URL limit, and every sitemap is still capped at 50,000 URLs or 50 MB uncompressed. Splitting sitemaps by language (sitemap-en.xml, sitemap-de.xml) under a sitemap index makes indexing problems in one language easy to spot in Search Console. Remember that a sitemap can only list URLs under the directory it is served from, so host language sitemaps at the root or in the matching folder.
Common mistakes checklist
Missing return links: one-way annotations are ignored.
Region-only or invented codes: "uk", "eu", "be" for Belgium, or es-419.
Cross-language canonicals: every translation canonicalized to English.
Relative URLs in hreflang values.
Alternates that redirect, 404 or are noindexed: point only at live, indexable, self-canonical URLs.
Automatic redirects by IP or browser language: Google advises against them because they can stop users and crawlers from reaching all versions; offer visible language links instead.
Mixing methods: HTML tags and sitemap entries that disagree after a content change.
Half-translated pages: translated navigation around untranslated main content.
For market selection before you translate anything, see strategies for reaching new markets, and for the tax and consumer rules that come with selling in other EU countries, selling across Europe. Page speed matters in every language, so keep an eye on Core Web Vitals as you add locales.
How CIQRA handles this
In CIQRA, per-language XML sitemaps, hreflang and canonical tags are generated automatically, and URL and meta templates are built in. The built-in AI writing tools cover SEO copy and translations, and when a site moves over, old URLs are kept or 301-redirected so existing rankings carry across. You still decide the language structure and check each translation; the platform keeps the annotations consistent.
FAQ
Does hreflang help a page rank higher?
No. It tells Google which existing version to show a given searcher. Each version still has to earn its ranking on its own content.
Should I use subdirectories or ccTLDs?
Subdirectories are easier to run on one host. ccTLDs give the clearest country signal but cost more and target a single country each.
Do I need x-default?
It is recommended rather than required. Add it whenever you have a language selector or a sensible fallback page for unmatched visitors.
Can hreflang point to another domain?
Yes. Alternate URLs do not need to be on the same domain, as long as the links are reciprocal.