Should you really self-manage canonicals on a multilingual site?

Why does Google emphasize self-referencing canonicals in an international context?

The canonical tag tells Google which URL is authoritative when multiple pages present similar content. On a monolingual site, pointing all variants to a main URL makes sense. However, on a multilingual site, each language version targets a distinct audience and must be indexed separately.

When a French page points its canonical to the English version, Google interprets: "The English page is the original, ignore the French one." As a result, the French version gradually disappears from the index or loses its ranking autonomy. The hreflang tags exist precisely to indicate to Google that these pages are language equivalents without creating a hierarchy among them.

How does the separation between canonical and hreflang actually work?

Each URL must self-manage: the /fr/ page has a canonical pointing to /fr/, the /de/ page to /de/, and so on. Hreflang tags, in turn, weave the network among these pages by specifying: "This page exists in German here, in Spanish there, in French elsewhere."

This allows Google to display the correct variant based on the researcher's language, without one version overshadowing another. It’s a decentralized architecture: each language is autonomous, linked to others by hreflang but never subordinate through canonical.

Does this rule also apply to regional variants of the same language?

Yes. A page in British English and one in American English should each have their own canonical, even if the text differs little. The regional variants (en-GB, en-US) target different audiences, sometimes with distinct terms, currencies, or date formats.

Forcing all variants to en-US would drain the others of their local relevance. The hreflang x-default allows you to designate a default version for users outside covered areas without creating canonical subordination.

• Mandatory self-canonical: each language version points its canonical tag to itself.
• Hreflang as a link: hreflang tags connect the variants without creating hierarchy.
• Independent indexing: each language retains its own ranking potential.
• No master version: no URL should dominate the others through cross-language canonical.
• Regional variants included: en-GB, en-US, es-MX, es-ES follow the same logic.

Is this statement consistent with field observations?

Yes, and it's one of the few Google guidelines that has no ambiguity. Audits regularly reveal sites where all versions point their canonical to en/, draining /fr/, /de/, and /es/ of all autonomy. These pages end up in indexing limbo: present in Search Console but invisible in local SERPs.

The correction immediately reverses the trend. Once canonicals are self-managed and hreflang correctly implemented, language versions rise within weeks. There's no technical haze here: the rule works exactly as Google announces.

What errors persist despite this clarity?

The most common: confusing canonical and hreflang x-default. Some practitioners point all canonicals to the x-default language, thinking they are "simplifying" the structure. This is a mistake. The x-default indicates a language redirect page or a default version for users outside areas, but never justifies canonical subordination.

Another trap: poorly configured CMS systems that automatically generate cross-language canonicals. Multilingual WordPress, Shopify Markets, and Magento multistore frequently fall into this default trap. Manual verification is required because a plugin can very well insert "smart" canonicals that sabotage everything.

In what cases could this rule be circumvented?

No legitimate cases. Even for identical content translated word for word, each language deserves its own indexing if it targets a distinct audience. The only exception: if two language versions strictly share the same URL (rare, but possible with parameters ?lang=), then one canonical suffices, and hreflang becomes unnecessary.

If you're unsure whether to merge or separate, ask yourself: "Do I want this page to rank in Google.fr AND Google.de?" Yes answer = separate canonicals. No answer = one URL suffices, and you don’t need hreflang. [To be verified] if you see contrary advice elsewhere, it’s outdated or erroneous.

What should you concretely do on an existing site?

Start with a canonical audit: export all URLs from your site using Screaming Frog or Sitebulb, filter by language, and check that each /fr/ points to /fr/, and each /de/ points to /de/. If a canonical points to another language, you’ve identified your problem.

Next, check the hreflang tags: each page must list all its language variants, including itself. A /fr/ page should contain hreflang x-default, hreflang en, hreflang de, hreflang fr. The absence of self-referenced hreflang is a common error that breaks the reciprocity required by Google.

How to fix a poorly structured canonical architecture?

If your canonicals all point to en/, you'll need to go through each template to enforce self-referencing. On WordPress, this might mean disabling Yoast or Rank Math for canonicals, and coding the tag manually. On Shopify, you may sometimes need to edit theme.liquid. For custom setups, modify the tag generator in the CMS.

After correction, monitor Search Console: previously ignored pages will reappear under "Indexed Pages." Expect 3 to 6 weeks for a full recrawl. If certain versions remain absent, ensure the XML sitemaps include all languages and that robots.txt is not blocking anything.

What errors to avoid during implementation?

Never mix canonicals and 301 redirects. If /fr/ redirects to /en/, there’s no need to discuss canonical or hreflang; the French page no longer exists. Geo-targeted redirects (based on IP) also pose issues: Googlebot crawls from the USA, will always see the US version, and ignore the others.

Another trap: hreflang tags in HTTP header versus HTML. Both work, but do not mix them on the same page. Choose one method and stick to it. Hreflang sitemaps are a viable option for very large sites, but more complex to maintain.

• Audit all canonicals: each language must point to itself
• Check the reciprocity of hreflang: each page must list all its variants, including itself
• Ensure that XML sitemaps include all language versions
• Test with a VPN or Search Console to validate display by geographic area
• Monitor "Excluded Pages" in GSC after correction to detect any setbacks
• Document the configuration in an internal wiki to avoid regressions during CMS migrations