Guide Multilingual SEO

TL;DR — Multilingual SEO is optimizing a site that’s disponible in plus que un language so the correct version montre up pour the correct person. It’s à propos de language, pas country — that’s the premier chose to obtenir straight. Vous give chaque language its propre URL, tell Google qui URLs are versions of chaque autre with hreflang, and let Google figure out the language from the réel text on the page.

Multilingual vs. multi-regional — obtenir ce straight premier

Ces two obtenir mixed up constantly, and the whole topic rend plus sense une fois vous separate les:

• Multilingual = votre site sert plus que un language. A Swiss site in German, French, and Italian is multilingual — three languages, un country.
• Multi-regional = votre site targets utilisateurs in plus que un country. A site with separate US and UK English versions is multi-regional — un language, two regions.

Plenty of sites are les deux (think a global brand with English, Spanish, and French, chaque split by country). But the signals and the problems are différent, so it helps to know qui un you’re en réalité dealing with. Ce page is à propos de the language side.

Ce que “language targeting” en réalité signifie

Quand votre site has the même page in several languages, vous vouloir Google to:

1. Know chaque language version exists.
2. Comprendre they’re versions of the même chose.
3. Montrer the German un to German searchers and the French un to French searchers.

Vous faire que se produire with a handful of signals — chiefly the hreflang attribute, qui is a little tag que dit “ce URL is the Spanish version, que one’s the French version, here’s the relationship entre les.” Google alors tries to serve the matching un.

The partie beginners are surprised by: Google figures out ce que language une page is in by reading the visible words on it — pas from quelconque setting in votre code. So a page que dit it’s German in le code but is complet of English text va be treated as English. Écrire contenu réel in the target language; that’s the signal que counts.

Où the différent language versions live (Structure d’URL)

Vous besoin a separate URL pour chaque language. Là are three courant façons to do it:

• Subdirectory — example.com/es/, example.com/fr/. The usual pick pour la plupart sites: facile to définir up, and tout the versions share the strength of votre un domain.
• Subdomain — es.example.com, fr.example.com. Fonctionne, but moteur de recherches treat it plus comme a separate site.
• Country-code domain (ccTLD) — example.es, example.fr. The strongest signal pour targeting a country, but it’s a whole separate website to construire authority pour.

Éviter stuffing the language into une URL parameter comme example.com?lang=es — Google specifically recommends contre it.

Quand vous en réalité besoin multilingual SEO (vs. simplement translating)

Si vous seulement translate the words and dump les on the même URL — or auto-redirect everyone by leur emplacement — vous don’t really have multilingual SEO, and Google may jamais même trouver votre autre languages. Vous besoin the réel chose (separate URLs + hreflang + content Google peut explorer) une fois you’re genuinely serving multiple languages and vouloir chaque un to rank. Si you’ve seulement got un language, vous pouvez skip tout of ce.

Vouloir the technical depth — how Google’s language algorithm fonctionne, the Bing differences, the courant hreflang errors, and the 2025 auto-translation threat? Switch to the Avancé tab.

Multilingual SEO mistakes

Treating lang as Google’s language detector

Pourquoi it fails: Google determines language from visible page text plutôt que relying on the HTML attribute. Do à la place: faire the rendered principal content consistently utiliser the target language; retain correct metadata pour accessibility and engines que utiliser it.

En utilisant hreflang to declare une page’s language

Pourquoi it fails: hreflang maps alternate URLs; it ne fait pas replace meaningful translated content. Do à la place: publish a distinct crawlable URL with visible target-language content, alors connecter equivalent versions with reciprocal hreflang.

Forcing visitors via an automatic language redirection

Pourquoi it fails: robots d’exploration and multilingual utilisateurs peut be trapped on the assumed version. Do à la place: garder every version directement accessible and offer a visible suggestion plus crawlable language liens.

Testez vos connaissances: Multilingual SEO

TL;DR — Obtenir the mental model correct and la plupart multilingual SEO falls into placer. Google fait pas lire votre <html lang> attribute or Content-Language HTTP header to detect language — it primarily uses visible text. Autre engines may utiliser différent signals, so vérifier leur current guidance. hreflang doesn’t declare une page’s language; it maps qui URLs are alternates so Google peut swap in the correct un — and it’s a hint, pas a directive. It’s aussi fiendishly error-prone: my study of 374 756 domains trouvé 67% had hreflang problèmes. Utiliser subdirectories pour la plupart sites, jamais auto-redirect Googlebot, and seulement translate pages with a réel business cas.

Multilingual ≠ multi-regional

The foundational distinction. Multilingual is à propos de language; multi-regional is à propos de country. A Swiss site in German, French, and Italian is multilingual but single-regional. A US/UK English site is multi-regional but single-language. Ils overlap tout the temps, but ils utiliser différent signals and créer différent problems, so I garder les separate in my head and so devrait vous. Google’s propre framing: “A multilingual website is quelconque website que offers content in plus que un language,” versus “A multi-regional website is un que explicitly targets utilisateurs in différent countries.”

How Google en réalité detects page language (ce is the big un)

Ici is the unique most-misunderstood chose in ce whole topic, and it’s worth overstating parce que so nombreux personnes obtenir it incorrect:

Google ignores votre <html lang> attribute. Google ignores votre Content-Language HTTP header. It reads the visible text.

Straight from the docs: “Google uses the visible content of votre page to determine its language. We don’t utiliser quelconque code-level language information tel as lang attributes, or l’URL.” And on the hreflang page: “Google doesn’t utiliser hreflang or the HTML lang attribute to detect the language of une page; à la place, we utiliser algorithms to determine the language.”

On Search Off the Record (July 2024), Gary Illyes gave ce a nom: LDCP — Language Demotion, Country Promotion. His explanation: “si someone is searching in German and votre page is in English, alors vous voudrait obtenir a demotion.” That’s a ranking mechanism, and it’s plus fundamental que hreflang — si votre page isn’t genuinely in the searcher’s language, LDCP demotes it regardless of ce que votre tags dire. (Illyes aussi signaled ce kind of manual signaling is on the façon out, noting Google déjà had systems “automatically learning” to pair alternate-language versions — auto-detection is the direction of travel.)

Pourquoi fait Google distrust <html lang> so beaucoup? Parce que it’s incorrect constantly. Gary told the Joomla story: the CMS “came with the Lang attribute définir to English, and là was aucun façon to override it… And alors vous looked at lune page, and it was 100% German. So it was comme vous pouvez’t trust que.” John Mueller has said the même — the HTML lang attribute is “something we don’t use at all” and “almost toujours incorrect. So we tend to ignore que.”

The takeaway: language detection is a content problem, pas a markup problem. Écrire contenu réel in the target language. The lang attribute encore matters pour accessibility (screen readers) and pour Bing — simplement pas pour Google’s language detection.

Bing fait it the opposite façon

Bing ne fait pas prise en charge hreflang. And où Google ignores votre code-level language hints, Bing dépend on les. Bing’s priority order:

1. <meta http-equiv="content-language" content="en-us"> — the principal signal.
2. <html lang=""> attribute — secondary.
3. On-page text content.

Bing’s format is a 2-letter ISO 639 language code, a dash, alors an ISO 3166 geography code (en-us, es-mx). So a correctement internationalized site running on les deux engines veut distinct-URL content plus hreflang (pour Google/Yandex) plus le contenu-language meta tag (pour Bing). They’re pas redundant; ils feed différent engines.

Ce que hreflang en réalité fait (and doesn’t)

hreflang is a relationship map, pas a language detector. It doesn’t tell Google the language of une page — Google déjà obtient que from content. It indique Google qui URLs are alternates of chaque autre and pour qui language-region audience, so Google peut swap in the correct un per searcher. Google: “Utiliser hreflang to tell Google à propos de the variations of votre content, so que we peut comprendre que ces pages are localized variations of the même content.”

Three choses have to be correct or the cluster is ignored:

• Bidirectional. “Si two pages don’t les deux point to chaque autre, the tags va be ignored.” English points to Italian, Italian doit point back.
• Self-referencing. Chaque page listes itself plus tout the others.
• x-default as a fallback pour unmatched languages (recommended, pas strictly requis).

Three delivery méthodes: HTML <link> tags in the <head>, the HTTP Link header (utile pour non-HTML fichiers comme PDFs), or an XML sitemap with the xhtml:link namespace. URLs doit be entièrement qualified notamment https://.

And critically: hreflang is a hint, pas a directive. Google’s propre line is que “hreflang annotations are a hint to Google, not a directive.” It peut and va override vous quand balise canonicals, content signals, or indexation status disagree.

The valuable, under-covered angle: hreflang speeds indexation of nouveau languages

Here’s a benefit almost nobody talks à propos de. On the même SOTR episode, Gary Illyes said discovering hreflang triggers a verification explorer of the alternates: “Là is a dependency explorer triggering quand we découvrir Hreflang, comme we vouloir to vérifier que.” In plain terms — quand Google trouve hreflang on une page, it goes and crawls the alternate-language versions to vérifier the relationship, qui speeds up indexation of votre nouveau language pages. So hreflang isn’t seulement à propos de serving the correct version; it’s a discovery accelerant pour newly launched languages.

Pourquoi votre GSC hreflang données semble thin (and that’s normal)

Un autre Illyes point worth internalizing: “Search Console seulement reports canonicals… the vast majority of the Hreflang clusters… ne sont pas canonical.” La plupart hreflang alternates aren’t URL canoniques, so ils don’t montrer up individually in GSC’s coverage reporting. Si votre dashboards faire it regarder comme votre language pages “aren’t indexed,” that’s souvent simplement GSC reporting canonicals — pas a réel problem. Vérifier with the Inspection d’URL outil and live SERPs plutôt que trusting the aggregate count.

Contenu dupliqué: es-MX vs. es-ES

A courant fear is que two Spanish variants va be flagged as duplicates. Google’s réel rule: “Localized versions of une page are seulement considéré duplicates si the principal content of lune page remains untranslated.” Two genuinely-Spanish pages with réel regional differences (currency, vocabulary, local références) are distinct. But two es-ES and es-MX URLs with identical text and aucun regional differentiation va be treated as duplicates — hreflang alone doesn’t separate les; distinct regional content fait. Dans a valid hreflang cluster, signals peut aussi consolidate à travers the alternates, qui is généralement utile.

Structure d’URL, maillage interne, and language detection

Structure. Subdirectories (/es/, /fr/) pour la plupart sites — ils inherit root domain authority and are low-maintenance. Subdomains obtenir treated plus comme separate sites. ccTLDs are the strongest country geotargeting signal but mean separate properties and separate link-building. URL parameters are explicitly “pas recommended.”

Maillage interne entre versions. Lien language versions to chaque autre sensibly (the language switcher is partie of ce), but don’t treat every alternate as a principal nav lien. Let hreflang carry the formal relationship; let the switcher serve utilisateurs.

Language detection UX — and the crawlability trap. Ce is où sites quietly break themselves. Googlebot “sends HTTP requêtes sans setting Accept-Language in la requête header” and crawls mostly from US IPs (though it fait utiliser some non-US IPs aussi). Si vous auto-redirect utilisateurs to a language version fondé on IP or Accept-Language, Googlebot may seulement ever voir votre English version and jamais découvrir the rest. Google’s blunt instruction: “Éviter automatically redirecting utilisateurs from un language version of a site to a différent language version.”

Do ce à la place: serve distinct URLs, annotate with hreflang, and si vous vouloir to aider utilisateurs, montrer a non-intrusive nudge banner (“Ce page is disponible in English — switch?”) que garde les on leur current URL and remembers leur choice via cookie or localStorage. Utiliser Accept-Language as a suggestion, jamais a redirection trigger. And don’t utiliser flags as language identifiers — flags are countries, pas languages.

Translation quality and traduction automatique

Traduction automatique itself isn’t banned. Google’s propre definition: “Scaled content abuse is quand nombreux pages are generated pour the principal objectif of manipulating search rankings and pas helping utilisateurs.” Jump to quote Translation is named as un exemple of an automated transformation covered by que policy — “notamment via automated transformations comme synonymizing, translating, or autre obfuscation techniques,” specifically “où little valeur is provided to utilisateurs.” Jump to quote The judgment is valeur, pas production méthode. Raw MT dumped at scale with aucun editorial oversight is the réel risk; MT with human examiner and curation (a workflow souvent appelé MTPE) is a standard, scalable façon to stay clair of it. Judge a translated page the façon you’d judge quelconque page — fait it serve réel search intent pour que market — and don’t simplement translate votre English keywords: do réel keyword research per language, parce que intent and phrasing differ market to market. (Plus depth: traduction automatique and SEO.)

Schema pour multilingual sites

Utiliser the inLanguage property on votre Article/Product/FAQPage schema to state the content’s language, and give chaque language version its propre schema block in que language. Schema and hreflang do différent jobs: hreflang maps alternate URLs; schema clarifies content meaning and language. They’re complementary layers.

The error rate: my 374 756-domain study

I ran a study à travers 374 756 domains en utilisant hreflang. The headline: over 67% had problèmes. The breakdown, la plupart courant premier: manquant x-default (56,3%), manquant self-referencing tags (18%), broken/redirigé références (16,9%), manquant reciprocal tags (15,3%), non-canonical références (8%), incorrect ISO codes (4,6%), inconsistent HTML lang vs. hreflang (3,2%), duplicate language entries (2,5%), and the même page declared pour multiple languages (2,5%).

Two caveats garder ce from being doom. Premier, the top two — x-default and self-referencing — are recommended, pas strictly requis, so the real-severity picture is softer que 67% suggests. Second, Google fixes a lot of ce pour vous. Gary Illyes has said hreflang errors “didn’t cause problems” in his implementation era, and my propre long-standing position (from my Enterprise SEO Chaos SMX talk) is que it may pas be worth votre temps chasing every hreflang error parce que Google is probably correcting nombreux of les automatically. But the structural errors — manquant reciprocal liens, broken URLs — do break the cluster and devrait be fixed. Triage: fix structural, monitor behavioral.

The 2025 threat: Google’s translate.goog proxy

A newer wrinkle worth knowing à propos de (Motoko Hunt’s June 2025 Moteur de recherche Land piece): quand Google determines there’s aucun high-quality, local-language content disponible, it auto-translates content via a translate.goog proxy. Le navigateur matters — Chrome utilisateurs obtenir routed to the publisher’s original URL (trafic preserved), but Safari/Firefox utilisateurs go via translate.goog (trafic lost). Ahrefs données put ce at 6,2 million AI Overview appearances with translated URLs and hundreds of millions of clicks of exposure. The strategic lire: publishing native local-language content is les deux the defense (Google won’t proxy-translate ce que vous déjà provide) and the offense (vous rank pour ceux markets yourself). Treat auto-translated SERPs as a signal of unmet demand.

Mueller’s caution: don’t translate everything

Finalement, restraint. John Mueller warns contre the all-languages reflex — “Let’s créer tout languages! Let’s faire pages pour tout countries! Ce que si someone in Japan veut to lire it in Swahili?” — and notes “it’s facile to dig into endless pits of complexity with hreflang.” His advice, which I echo: “premier… limite the number of pages vous créer to ceux que are absolutely critical & valuable.” Prioritize high-traffic, high-conversion pages in markets with demonstrated demand (utiliser GSC to trouver requêtes déjà pulling impressions in a target language). Every language version vous créer is explorer, indexation, canonicalization, and maintenance overhead — seulement spend it où there’s a réel retourner.

AI summary

A condensed prendre on the Avancé version:

• Multilingual ≠ multi-regional. Multilingual is à propos de language (Swiss site in 3 languages, 1 country); multi-regional is à propos de country (US/UK English). Ils overlap but utiliser différent signals.
• Google detects language from visible text — pas <html lang>, pas the Content-Language header, pas l’URL. Gary Illyes named the ranking mechanism LDCP (Language Demotion, Country Promotion): wrong-language pages obtenir demoted.
• Bing is the opposite: it doesn’t prise en charge hreflang and uses <meta http-equiv="content-language"> as its principal signal (priority: content-language meta > html lang > content).
• hreflang is a relationship map, pas a language detector, and it’s a hint, pas a directive. Doit be bidirectional + self-referencing; x-default recommended.
• Bonus benefit: discovering hreflang triggers a verification explorer, speeding indexation of nouveau language pages (Illyes). La plupart alternates aren’t canonical, so GSC under-reports les — that’s normal.
• es-MX vs es-ES are seulement duplicates si the principal content is untranslated; genuine regional differences separate les.
• Structure d’URL: subdirectories pour la plupart sites; ccTLDs strongest pour country targeting; éviter URL parameters. Jamais auto-redirect by IP/Accept-Language — Googlebot crawls sans Accept-Language and mostly from US IPs, so it may miss votre autre languages.
• Traduction automatique isn’t banned. Google’s scaled-content-abuse policy noms translation as un exemple of an “automated transformation,” but the violation is publishing nombreux low-value pages to manipulate rankings — the judgment is valeur, pas méthode. Raw MT dumped at scale with aucun editorial oversight is the réel risk; MT with human examiner (MTPE) is a standard, scalable workflow.
• Error rate: my study of 374 756 domains trouvé 67% had hreflang problèmes, but Google auto-corrects nombreux — fix structural errors, monitor the rest.
• 2025 threat: Google’s translate.goog proxy peut hijack trafic où native content is manquant; native local content is les deux defense and offense.

Documentation officielle

Primary-source documentation from the moteur de recherches.

Google

• Managing Multi-Regional and Multilingual Sites — the foundational doc: definitions, Structure d’URL options, locale signals, the auto-redirect warning.
• Localized Versions of votre pages — the hreflang référence: the three implementation méthodes, bidirectionality, self-referencing, x-default, and the duplicate-content rule.
• International SEO Overview — the entry point pour language/country/region optimization.
• How Google Crawls Locale-Adaptive Pages — pourquoi serving différent content by perceived locale risks explorer/index gaps, and the Accept-Language remarque.

Bing / Microsoft

• How To Tell Bing Votre Website’s Country and Language — the content-language meta tag, its format, and the priority order Bing uses (encore the principal technical référence).
• Bing Webmaster Guidelines — the general guidelines, notamment language/country targeting.

Quotes from the source

On-the-record statements from Google and Bing on multilingual SEO.

Google — how language is detected

• “Google uses the visible content of your page to determine its language. We don’t use any code-level language information such as lang attributes, or the URL.” — Recherche Google Central docs. Jump to quote
• “Google doesn’t use hreflang or the HTML lang attribute to detect the language of a page; instead, we use algorithms to determine the language.” Jump to quote

Google — hreflang and duplicates

• “Use hreflang to tell Google about the variations of your content, so that we can understand that these pages are localized variations of the same content.” Jump to quote
• “If two pages don’t both point to each other, the tags will be ignored.” Jump to quote
• “Localized versions of a page are only considered duplicates if the main content of the page remains untranslated.” Jump to quote

Google — redirections and definitions

• “Avoid automatically redirecting users from one language version of a site to a different language version.” Jump to quote
• “A multilingual website is any website that offers content in more than one language.” Jump to quote

Gary Illyes, Google (Search Off the Record, “Internationalization & hreflang,” July 25, 2024)

• On the LDCP algorithm: “if someone is searching in German and your page is in English, then you would get a demotion.”
• On pourquoi HTML lang is unreliable: “[Joomla] came with the Lang attribute set to English, and there was no way to override it… And then you looked at the page, and it was 100% German. So it was like you can’t trust that.”
• On hreflang triggering a verification explorer: “There is a dependency crawl triggering when we discover Hreflang, like we want to verify that.”
• On GSC reporting: “Search Console only reports canonicals… the vast majority of the Hreflang clusters… are not canonical.” Listen to the episode

John Mueller, Google

• On the HTML lang attribute: “the language attribute within the HTML markup is something we don’t use at all” and “almost always wrong. So we tend to ignore that.”
• On over-creating languages: “It’s easy to dig into endless pits of complexity with hreflang.” / “My recommendation would be first to limit the number of pages you create to those that are absolutely critical & valuable.”

The decision frameworks

1. Do vous besoin multilingual or multi-regional? (or les deux)

Run ce avant anything sinon — it determines qui signals vous reach pour:

Are you serving more than one LANGUAGE?
├─ No  → not a multilingual problem. Stop here for language work.
└─ Yes → multilingual. You need: distinct URLs per language
         + hreflang (Google/Yandex) + content-language meta (Bing)
         + real content in each language.

Are you also targeting more than one COUNTRY?
├─ No  → single-region multilingual (e.g. Swiss DE/FR/IT).
│        Use language-only hreflang codes (es, fr, de).
└─ Yes → multi-regional too. Add region to hreflang (es-MX, es-ES,
         en-GB, en-US) and consider ccTLDs for the strongest
         country signal.

The trap: treating language as a country signal. As Gary Illyes put it, “the language is absolutely not a tell for what country you are targeting.”

2. Structure d’URL comparison

3. Content priority matrix — ce que to translate premier

Don’t translate everything. Score candidate pages and fonctionner top-down:

Priority = (organic demand in target market)
         × (conversion value of the page)
         ÷ (translation + maintenance effort)

Practical order:

1. Proven-demand pages — utiliser GSC to trouver URLs déjà getting impressions/clicks from a target-language requête or country. Demand is déjà là; capture it.
2. High-traffic / high-conversion core pages — money pages and top cluster hubs.
3. Supporting cluster pages — une fois the core market is validated.
4. Long-tail — dernier, and seulement si the market is paying off.

Mueller’s rule sits on top of tout of it: “limite the number of pages vous créer to ceux que are absolutely critical & valuable.” Every supplémentaire language version is ongoing explorer, indexation, and maintenance cost — spend it où there’s retourner.

Language signals — ce que chaque engine en réalité uses

The un table to bookmark. Google and Bing lire almost opposite signals.

Bing priority order: content-language meta tag > <html lang> > page content.

Google in un line: it reads the visible text to detect language; everything code-level (lang, Content-Language, URL) is ignored pour que objectif.

hreflang non-negotiables: bidirectional, self-referencing, fully-qualified URLs with https://, x-default recommended. Hint, pas directive.

Don’t: auto-redirect by IP/Accept-Language (Googlebot crawls sans Accept-Language, mostly from US IPs — it’ll miss votre autre languages); utiliser flags pour languages; dump raw, unreviewed traduction automatique at scale with aucun editorial value-add.

Journal des modifications