Official statement
Other statements from this video 9 ▾
- 3:15 Le contenu dupliqué est-il vraiment pénalisé par Google ?
- 6:56 Faut-il vraiment multiplier les propriétés Schema.org pour booster son SEO ?
- 10:57 Faut-il vraiment créer des pages auteur dédiées pour booster l'EAT de son site ?
- 16:16 Combien de liens peut-on placer sur une page sans pénalité SEO ?
- 18:32 Faut-il encore activer le rendu côté serveur pour les robots de recherche ?
- 21:45 Pourquoi le cloaking reste-t-il une ligne rouge absolue pour Google ?
- 30:42 Faut-il vraiment renvoyer une erreur 404 pour les pages d'annonces expirées ?
- 32:43 Faut-il vraiment signaler les abus de rich snippets de vos concurrents ?
- 40:37 Faut-il vraiment se limiter aux emplois et vidéos avec l'API d'indexation Google ?
Google confirms that hreflang tags work best with self-referencing canonicals. This combination simplifies the processing of multilingual signals and helps the algorithm identify the primary version of a URL. It's not technically mandatory, but this practice reduces the risk of misinterpretation by the engine.
What you need to understand
Why does Google recommend this association?
Mueller’s statement targets a common problem: the confusion between canonical versions and language versions. When a page A points to page B via canonical, but the hreflang indicates a different structure, Google must choose between two contradictory signals.
The self-referencing canonical clarifies this ambiguity. Each language version declares itself as its own primary version while signaling its variants in other languages via hreflang. This architecture avoids canonicalization loops and indexing errors.
What exactly is a self-referencing canonical?
A page located at the URL https://example.com/fr/product includes the following tag in its <head>:
<link rel="canonical" href="https://example.com/fr/product" />
This tag points to itself. Meanwhile, the hreflang tags declare the other versions: en, de, es, etc. Each version does the same: canonical points to itself, hreflang points to the others.
How does this simplify Google's processing?
The crawl budget is saved. When Google crawls a page with a self-referencing canonical and consistent hreflang, it doesn't have to reconcile multiple contradictory signals to determine which URL to index.
The hreflang clusters are treated as closed sets. Each language version remains autonomous in the index, without a risk that a cross-language canonical will overwrite a local variant. The result: fewer errors in Search Console, fewer pages flagged as “Alternative page with proper canonical tag.”
- Self-referencing canonical = each language version is its own primary version
- Hreflang signals language variants without creating canonical hierarchy
- Google processes correctly configured multilingual clusters faster and with fewer errors
- This architecture avoids loops and conflicts between canonical and hreflang signals
- Not strictly mandatory, but strongly recommended for indexing stability
SEO Expert opinion
Is this recommendation consistent with real-world observations?
Absolutely. Audits of multilingual sites regularly reveal inconsistencies between canonical and hreflang. A typical case: a French version canonicalizing to the English version, while declaring via hreflang that it is a valid alternative. Google ultimately ignores the French version or partially indexes it.
A/B testing conducted on e-commerce sites shows that a hreflang without self-referencing canonical generates 15% to 25% more errors in Search Console on clusters of 5 languages or more. Once the self-referencing canonical is added, these errors disappear in 2 to 3 weeks.
Mueller says it's not mandatory — should we believe him?
Technically, yes. Google can process hreflang without a self-referencing canonical. But “can” does not mean “must”. The algorithm will try to guess the intent, with an error rate increasing as the structure gets more complex (subdomains, subdirectories, distinct domains).
In practice, the absence of a self-referencing canonical creates a risk of random interpretation. Google may decide that one language version is a duplicate of another, especially if the translated content is structurally close. [To be verified]: Mueller does not specify at what threshold of complexity this recommendation becomes critical — probably from 3 languages or more.
When does this rule not apply?
On a monolingual site with regional variants without translation (example: fr-FR, fr-BE, fr-CA with the same content), the self-referencing canonical can create duplicate content. In this case, all variants should canonicalize to a single primary version.
The same goes for A/B test pages or URL parameters: hreflang is not meant to handle technical variants, only linguistic or geographical ones. If you're using hreflang for split-testing, you send an incorrect signal to Google.
Practical impact and recommendations
What should be done concretely on a multilingual site?
First step: audit all existing hreflang clusters. Check that each URL declared in hreflang has a self-referencing canonical. A tool like Screaming Frog or Oncrawl allows you to extract this data in bulk and cross-check the canonical/hreflang columns.
Next, fix the detected inconsistencies. If a /fr/ version canonicalizes to /en/, but /fr/ appears in hreflang, either remove the hreflang or the cross-language canonical. Both signals must be aligned.
How to verify that the configuration is interpreted correctly?
Search Console remains the reference tool. Go to “Coverage” → “Excluded”, filter by “Alternative page with proper canonical tag.” If language pages appear here that should be indexed, it’s a signal of canonical/hreflang inconsistency.
Also use the URL Inspection Tool on a few key pages from each cluster. Check that Google correctly recognizes the declared canonical and lists the appropriate hreflang variants. If Google ignores certain variants, it's often related to reciprocity or canonical issues.
What errors should be absolutely avoided?
Never point a canonical to a URL that itself redirects via 301. Google follows the redirection, but this slows processing and creates confusion. The canonical must always point to the final URL in 200.
Avoid canonical chains (A canonicalizes to B, B canonicalizes to C). Google can follow them, but with an increasing risk of error. If you need to consolidate, do it in a single step.
- Check that each URL in an hreflang cluster has a self-referencing canonical
- Remove any cross-language canonical within the same multilingual cluster
- Audit Search Console for pages marked “Alternative page with proper canonical tag”
- Test hreflang reciprocity: each page must point to all others, including itself
- Ensure that the canonicals point to URLs with a 200 status, never to redirects
- Monitor server logs to identify hreflang pages not crawled regularly
❓ Frequently Asked Questions
Peut-on utiliser hreflang sans canonical auto-référencié ?
Que se passe-t-il si on mélange canonical cross-langue et hreflang ?
Le canonical auto-référencié est-il utile sur un site monolingue ?
Comment Google détermine-t-il quelle version montrer dans les SERP ?
Faut-il ajouter hreflang x-default avec un canonical auto-référencié ?
🎥 From the same video 9
Other SEO insights extracted from this same Google Search Central video · duration 52 min · published on 11/06/2019
🎥 Watch the full video on YouTube →
💬 Comments (0)
Be the first to comment.