Official statement
Other statements from this video 27 ▾
- 13:31 Vos pages lentes peuvent-elles plomber le classement de tout votre site ?
- 13:33 Les Core Web Vitals impactent-ils vraiment tout votre site ou seulement vos pages lentes ?
- 13:33 Peut-on bloquer la collecte des Core Web Vitals avec robots.txt ou noindex ?
- 14:54 Pourquoi CrUX collecte vos Core Web Vitals même si vous bloquez Googlebot ?
- 15:50 Page Experience : Google ment-il sur son véritable poids dans le classement ?
- 16:36 L'expérience de page est-elle vraiment un signal de classement secondaire ?
- 17:28 Le LCP mesure-t-il vraiment la vitesse perçue par l'utilisateur ?
- 19:57 Les Core Web Vitals se calculent-ils vraiment pendant toute la navigation ?
- 20:04 Les Core Web Vitals évoluent-ils vraiment après le chargement initial de la page ?
- 21:22 Comment Google estime-t-il vos Core Web Vitals quand les données CrUX manquent ?
- 22:22 Comment Google estime-t-il les Core Web Vitals d'une page sans données CrUX ?
- 27:07 Comment Google attribue-t-il désormais les données CrUX du cache AMP à l'origine ?
- 29:47 AMP est-il encore nécessaire pour ranker dans Top Stories sur mobile ?
- 32:31 Comment exploiter les logs serveur pour détecter les erreurs 4xx dans Search Console ?
- 34:34 Pourquoi les nouveaux sites connaissent-ils une volatilité extrême dans l'indexation et le classement ?
- 34:34 Faut-il vraiment analyser les logs serveur pour diagnostiquer les erreurs 4xx dans Search Console ?
- 34:34 Pourquoi votre nouveau site fluctue-t-il comme un yoyo dans les SERP ?
- 40:03 Faut-il vraiment signaler le contenu copié de votre site via le formulaire spam de Google ?
- 40:20 Comment signaler efficacement le spam de contenu copié à Google ?
- 43:43 Vos pages franchise sont-elles des doorway pages aux yeux de Google ?
- 45:46 Le contenu dupliqué est-il vraiment sans danger pour votre référencement ?
- 45:46 Le contenu dupliqué est-il vraiment sans pénalité pour votre SEO ?
- 45:46 Vos pages franchises sont-elles perçues comme des doorway pages par Google ?
- 51:52 Le namespace http:// ou https:// dans un sitemap XML influence-t-il vraiment le crawl ?
- 52:00 Le namespace en https dans votre sitemap XML pénalise-t-il votre référencement ?
- 56:00 Faut-il vraiment soumettre les versions mobile ET desktop dans votre sitemap ?
- 61:54 Faut-il abandonner AMP si vous utilisez GA4 pour mesurer vos performances ?
Google confirms that a sitemap can contain only one version of the URL (mobile OR desktop) if the bidirectional annotations rel="alternate" and rel="canonical" are correctly implemented in the HTML code. This statement simplifies the technical management of sites with separate URLs, but relies on a critical condition: the accuracy of the annotations. In practice, an incomplete sitemap combined with faulty annotations can create serious indexing issues.
What you need to understand
Why does Google allow only submitting one version of the URL?
When a site maintains distinct URLs for mobile and desktop, Google relies primarily on the HTML annotations to understand the relationship between these versions. The rel="alternate" media="only screen and (max-width: 640px)" tags on the desktop side and rel="canonical" on the mobile side explicitly indicate to the crawler which version to serve based on the context.
If these annotations are present and consistent, Googlebot can discover all URL pairs by crawling only one version. The sitemap then becomes redundant for this specific function — it is merely used to speed up initial discovery or signal deep pages.
What constitutes a correct bidirectional annotation?
The bidirectional annotation requires that each URL points to its counterpart and vice versa. On the desktop version https://example.com/page, there should be a link to https://m.example.com/page with rel="alternate" and the appropriate media attribute. On the mobile version, a rel="canonical" must point back to the desktop version.
This symmetry allows Google to automatically map all pairs without needing a comprehensive inventory in the sitemap. But beware: a missing, malformed, or inconsistent annotation breaks this logic and can leave orphan pages or create canonicalization conflicts.
In what context does this practice still apply?
This setup primarily concerns sites that have chosen to maintain separate mobile/desktop URLs rather than adopt a responsive design with unique URLs. This architectural choice has become a minority since the widespread adoption of responsive design and mobile-first indexing.
The sites still affected are often historical platforms (legacy e-commerce, complex media portals) where a complete redesign is postponed for budgetary or technical reasons. For these sites, managing annotations and sitemaps remains critical, as an error can fragment indexing between the two versions.
- A sitemap can contain only one version (mobile OR desktop) if the HTML annotations are perfectly implemented.
- Bidirectional annotations (rel="alternate" + rel="canonical") must cover 100% of URL pairs without exception.
- This approach pertains to sites with separate URLs, a minority configuration compared to responsive design.
- The sitemap remains useful for rapid discovery and managing deep pages, even with correct annotations.
- Google uses HTML annotations as the primary source for mapping mobile/desktop relationships.
SEO Expert opinion
Is this statement consistent with real-world observations?
On paper, Google's logic is perfectly defensible. HTML annotations are indeed more reliable than a sitemap for defining canonical relationships, as they are directly attached to each resource. In theory, a crawler can rebuild the entire tree by following these links.
However — and here lies the problem — in practice, annotations are rarely perfect across an entire site. Field audits regularly reveal inconsistencies: orphan pages without rel="canonical", asymmetric annotations (present in one direction but not the other), miscalibrated media attributes, or worse, conflicts where multiple mobile pages point to the same desktop page with different canonicals. [To be verified] to what extent Googlebot tolerates these partial errors without degrading indexing.
What risks does this minimalist approach entail?
Including just one version in the sitemap creates a total dependency on the quality of the annotations. If an annotation is missing or erroneous in one part of the site, Google may never discover the counterpart mobile (or desktop) pages. The risk is particularly high on large sites with dynamically generated content.
Some practitioners prefer a defensive approach: submit both versions in the sitemap, even if annotations are present. This redundancy provides an extra layer of security and speeds up discovery, especially after a massive deployment or update. Google does not penalize this practice — it simply ignores detected duplicates.
Should we still care about this configuration in a mobile-first indexing environment?
Since the widespread shift to mobile-first indexing, Google crawls and indexes the mobile version primarily. For a site with separate URLs, this means the mobile URL effectively becomes the canonical reference — even if the rel="canonical" technically points to the desktop.
This inversion sometimes creates confusion: SEO professionals observe that Google indexes the mobile version while displaying the desktop URL in the SERP, thanks to the annotations. As a result, the mobile sitemap becomes more critical than the desktop. If you could submit only one version, favor the mobile — this is what Googlebot crawls first and what serves as the basis for content evaluation.
Practical impact and recommendations
What steps should be taken on a site with separate URLs?
Start with a complete audit of bidirectional annotations. Export all mobile/desktop URL pairs from your CMS or database, then check that each desktop page contains a rel="alternate" pointing to its mobile version, and vice versa a rel="canonical" from the mobile to the desktop. Tools like Screaming Frog or OnCrawl can automate this verification on scale.
If the annotations are solid and comprehensive, you can technically submit only one sitemap — favor the mobile sitemap, as this is the version that Googlebot crawls first since the mobile-first indexing. But monitor coverage in Search Console: any absent or excluded page should trigger an immediate check of the corresponding annotations.
What mistakes to avoid in this configuration?
The classic mistake is to correct annotations on only part of the site, creating an inconsistency between sections. For example, new pages follow best practices, but the old catalog remains orphaned. Google may then index partially, with mobile or desktop versions missing based on section.
Another pitfall: changing URLs (structure change, moving from /m/ to m.example.com) without updating existing annotations. Old pages continue pointing to outdated URLs, breaking bidirectionality. An incomplete sitemap then hides these errors until they translate to traffic drops.
How to check if the configuration is working correctly?
Use Search Console to compare submitted URLs and indexed URLs. If you submitted only the mobile sitemap, check that desktop versions correctly appear in the coverage reports with a status of "Alternative page with proper canonical tag". If desktop pages are indexed when you intended to prioritize mobile, this is a signal of faulty annotations.
Also test using the URL inspection tool: enter a mobile URL and check that Google correctly recognizes the canonical pointing to the desktop (and vice versa). The inspection report explicitly displays the declared canonical and the one Google has selected — any discrepancies warrant investigation.
- Audit the completeness and symmetry of rel="alternate" and rel="canonical" annotations across the site.
- Favor the mobile sitemap if you submit only one version, as Google crawls in mobile-first.
- Monitor Search Console coverage reports to detect orphan pages or canonicalization conflicts.
- Set up automated annotation monitoring with every deployment of new pages.
- Regularly test with the URL inspection tool to validate that Google correctly interprets mobile/desktop relationships.
- Document the sitemap strategy to avoid regressions during team or vendor changes.
❓ Frequently Asked Questions
Dois-je absolument soumettre les deux versions mobile et desktop dans mon sitemap ?
Quelle version du sitemap privilégier si je n'en soumets qu'une ?
Comment vérifier que mes annotations bidirectionnelles sont correctes ?
Que se passe-t-il si une annotation est manquante sur certaines pages ?
Cette recommandation s'applique-t-elle aux sites responsive avec URLs uniques ?
🎥 From the same video 27
Other SEO insights extracted from this same Google Search Central video · duration 1h07 · published on 28/01/2021
🎥 Watch the full video on YouTube →
💬 Comments (0)
Be the first to comment.