Does each language version really need its own self-referencing canonical?

Why does this rule contradict what many SEOs apply?

For years, some practitioners have used the canonical tag to "consolidate" language versions deemed "secondary" to a primary language — often English. The idea was to concentrate the SEO juice on a single URL and avoid the perceived dilution of crawl budget.

However, Google does not operate this way. Mueller is very clear: if you point the canonical of your French version to your English version, Google interprets this as you explicitly saying, "ignore the French version, only process the English version." As a result, the French version disappears from the index. This is not a bug; it's the expected behavior of the engine.

What is the difference with hreflang then?

The confusion often arises from mixing canonical and hreflang. The two tags serve completely different purposes. The canonical indicates which URL should be indexed among several URLs with identical or very similar content. The hreflang tells which language or regional version to display to which user.

In short: you use hreflang to inform Google "here is the English version for English speakers, the French version for French speakers." And you use the canonical to specify, "this FR page is indeed the version to index for this FR URL." The two tags work in tandem, not one instead of the other.

What happens concretely if my canonical points to another language?

Google will treat the URL pointed to by the canonical as the reference URL. All other language versions pointing to this same canonical will be considered unwanted duplicates. They will not be indexed — or indexed randomly and unstably.

In the Search Console, you will see these pages in the "Excluded" tab with the status "Duplicate, page not selected as canonical." French-speaking users will never find your French version in Google.fr results, as only the English version will be served — with all the disastrous consequences this implies for the bounce rate and conversions.

• Each language version must have a self-referencing canonical pointing to itself
• The canonical does not replace hreflang — both tags are complementary and must coexist
• A cross-language canonical leads to the de-indexing of the "secondary" version
• Non-indexed URLs appear in the Search Console as excluded duplicates
• This error directly impacts the organic traffic of the affected language versions

Is this statement consistent with what we observe in the field?

Yes, absolutely. We regularly see international sites lose whole segments of their organic traffic on language versions because a developer misinterpreted the canonical tag. The problem is that the error is not always immediately visible — it can take weeks to materialize while Google recrawls and reevaluates the canonicals.

What's interesting is that Mueller does not say, "it's a bug" or "we're going to fix that." He explicitly states that it is the expected behavior. In other words: if you mess up your configuration, Google will not guess your intention. It will blindly apply what you tell it to do.

What nuances need to be added to this rule?

There is an edge case: sites that have strictly identical language versions in content, with only minor variations in translation. In this case, some SEOs deliberately choose not to index certain versions to avoid diluting their authority. But beware — this is a deliberate strategic choice, not a configuration error.

In this specific case, it is better to noindex the secondary versions outright rather than playing with the canonical. This way, there is no ambiguity: Google knows that you do not want to index these pages. With a cross-language canonical, you create a gray area where Google may or may not index according to its mood — and that's unmanageable.

Are there situations where this rule does not apply?

No. Let's be frank: this rule applies to 100% of multilingual sites. The only exception would be a site where language versions are not truly translations but radically different content — but then we are not truly in the multilingual realm.

What is missing from Mueller's statement is a clarification on sites with regional variations of the same language (fr-FR, fr-CA, fr-BE). Logically, each variation should also have its own self-referencing canonical, but some SEOs consolidate these variants into a single "generic" URL. [To be verified] — Google has never made a clear statement on this specific point.

How can I quickly audit the canonical configuration of my multilingual site?

Start with a Screaming Frog or Oncrawl crawl targeting all your language versions. Export the column "Canonical Link Element 1" and check that each URL points correctly to itself. If you see FR URLs pointing to EN URLs (or vice versa), you have a problem.

Next, go to the Search Console and filter excluded pages by "Duplicate, page not selected as canonical." If you see URLs of language versions there, it's a sign that Google has detected cross-language canonicals and decided not to index those pages.

What technical errors most often cause this problem?

The classic error: a CMS configured with a "default" language that serves as canonical for all others. This happens a lot on Magento, Shopify, or poorly configured custom CMS. The developer thinks they are doing the right thing by "consolidating" signals, but they sabotage indexing.

Another frequent case: sites that migrated from a multilingual structure in subdomains (fr.example.com) to subdirectories (/fr/) without updating the canonicals. The old canonicals still point to the old subdomains, creating an inconsistency that Google cannot resolve.

What should I do if my site already has this error in production?

Immediately correct the canonicals to point to themselves across all language versions. Then, force a recrawl via the Search Console by submitting the XML sitemaps for each language version. It can take a few weeks for Google to reindex the corrected pages.

Monitor the evolution in the Search Console: the number of excluded pages should decrease, and the number of indexed pages should increase. If after a month you see no change, check that you don't have other conflicting signals — notably unintentional noindex tags or forgotten 301 redirects.

• Crawl the entire multilingual site and export the canonicals for verification
• Check the Search Console for excluded pages due to cross-language canonicals
• Correct the canonicals to systematically point to themselves
• Submit the XML sitemaps for each language version to speed up recrawl
• Test hreflang tags with the Google validator to avoid conflicts
• Monitor indexing trends over 4 to 6 weeks following correction