What does Google say about SEO? /
Quick SEO Quiz

Test your SEO knowledge in 3 questions

Less than 30 seconds. Find out how much you really know about Google search.

🕒 ~30s 🎯 3 questions 📚 SEO Google

Official statement

Google takes multiple signals into account beyond the canonical tag to determine the canonical page. It may happen that Google still indexes variants despite the canonical, especially during the first crawl. The noindex tag allows you to enforce the selection of the canonical page.
272:01
🎥 Source video

Extracted from a Google Search Central video

⏱ 932h29 💬 EN 📅 05/03/2021 ✂ 13 statements
Watch on YouTube (272:01) →
Other statements from this video 12
  1. 9:53 Faut-il vraiment ignorer Schema.org pour les variantes de produits e-commerce ?
  2. 50:33 Pourquoi vos données structurées sabotent-elles votre Knowledge Panel ?
  3. 260:39 Le noindex des variantes produit contamine-t-il vraiment la page canonique ?
  4. 409:18 Comment Google évalue-t-il vraiment les Core Web Vitals d'une page dans ses résultats de recherche ?
  5. 434:38 La pertinence l'emporte-t-elle vraiment sur les Core Web Vitals dans Google ?
  6. 540:44 Faut-il vraiment maintenir les redirections 301 pendant un an minimum ?
  7. 595:13 Faut-il vraiment implémenter hreflang dès le lancement d'un site multi-pays avec contenu similaire ?
  8. 614:30 Pourquoi le linking interne entre versions linguistiques accélère-t-il vraiment l'indexation d'un nouveau marché ?
  9. 647:54 Faut-il vraiment doubler hreflang avec du JavaScript pour la géolocalisation ?
  10. 693:12 Pourquoi Google met-il plusieurs mois à récompenser les améliorations qualité d'un site ?
  11. 856:03 Faut-il s'inquiéter d'avoir 90% de pages en noindex sur son site ?
  12. 873:31 Faut-il vraiment utiliser un code 410 plutôt qu'un 404 pour supprimer une page de l'index Google ?
📅
Official statement from (5 years ago)
TL;DR

Google does not rely solely on the canonical tag to determine which page to index. Other signals come into play, which explains why variants can sometimes be indexed despite a well-configured canonical tag. If you want to force a variant not to be indexed, the noindex tag is necessary, not the canonical, which remains just one signal among many.

What you need to understand

Why does Google sometimes ignore my canonical tag?<\/h3>

The canonical tag<\/strong> is not an absolute directive. It is a signal among others<\/strong> that Google interprets to decide which version of a page to display in its results. In practice, Google cross-references this signal with dozens of other indicators: the consistency of internal linking, external backlinks pointing to one variant or another, the presence of conflicting hreflang tags, or even the freshness of content.<\/p>

During the first crawl<\/strong> of a page, Google may temporarily index several variants, while it analyzes these signals. This is particularly noticeable on e-commerce sites with multiple URL parameters (sorting, filters, pagination). You might have configured a proper canonical, but if Google detects that the variant with the parameter receives external links or generates user sessions, it may still decide to index it.<\/p>

What are the key signals that Google takes into account?<\/h3>

Google never shares an exhaustive list or their exact weighting, but field observations converge on several key elements. The internal linking<\/strong> plays a role: if you link internally to the variant rather than to the canonical, you confuse the signal. Backlinks<\/strong> carry significant weight — a variant that captures natural links will tend to be indexed, canonical or not.<\/p>

Redirects<\/strong> also come into play. If you change the canonical without implementing a 301 redirect from the old version, Google may hesitate. The consistency of content<\/strong> between variants also matters: two almost identical pages with a canonical will behave differently than two pages with divergent content. Finally, the HTTP response<\/strong> (200 vs 404 vs 410) and the presence or absence of hreflang tags can influence the final choice.<\/p>

When is the canonical really not enough?<\/h3>

Let’s be honest: if your goal is to guarantee the non-indexing<\/strong> of a page, the canonical does not guarantee that. Google can decide to override it if the conflicting signals are too strong. This is especially true on high-volume sites where crawling is fragmented, and signals update asynchronously.<\/p>

The noindex<\/strong> tag is the only directive that truly forces Google’s hand. Unlike the canonical, it prevents any indexing — but be careful, it also blocks the passage of PageRank. This is the critical nuance: canonical = signal consolidation, noindex = strict blocking. If you want to prevent the indexing of filter or pagination pages while preserving the flow of link equity to the consolidated page, you may find yourself in a technical deadlock that needs to be resolved on a case-by-case basis.<\/p>

  • The canonical is a signal, not a directive<\/strong>: Google can ignore it if other indicators point to a different variant.<\/li>
  • Backlinks and internal linking<\/strong> can contradict your canonical and force the indexing of an unwanted variant.<\/li>
  • Noindex is the only guarantee<\/strong> of non-indexing, but it cuts off PageRank flow — use it with caution.<\/li>
  • The first crawl<\/strong> may temporarily index several variants while Google consolidates its signals.<\/li>
  • The overall consistency<\/strong> of the site (redirects, hreflang, URL structure) strongly influences the consideration of the canonical.<\/li><\/ul>

SEO Expert opinion

Is this statement consistent with what we observe on the ground?<\/h3>

Yes, and it confirms what SEOs have noticed for years on high-traffic sites. E-commerce sites<\/strong> with filtering facets, advertisement sites<\/strong> with dynamic sorting, or editorial sites<\/strong> with complex pagination regularly encounter this phenomenon: variants are indexed despite a well-configured canonical. Google itself has reiterated this in various forms, but this wording points out the real issue: the canonical is a signal among<\/em> others.<\/p>

What is less commonly said is that the weighting of these signals varies<\/strong> depending on the context. On a new site with few backlinks, the canonical will probably be respected to the letter. On an older site with a complex link profile, Google may give more weight to the backlinks pointing to a variant than to your canonical. [To be verified]<\/strong>: Google has never published quantitative data on this weighting, and it’s likely that it is dynamic based on the site and the industry.<\/p>

What nuances should we add to Google's position?<\/h3>

The official discourse remains deliberately vague on when<\/strong> exactly Google chooses to override a canonical. The phrase "during the first crawl" implies that it’s temporary, but in practice, some variants remain indexed for months, or even indefinitely, even after multiple crawls. This is particularly noticeable on sites with active URL parameters<\/strong> (e.g., ?sort=price&filter=color) that generate long-tail organic traffic.<\/p>

Another rarely mentioned point: the canonical can be ignored if Google detects manipulation<\/strong>. For example, canonicalizing from page A to a radically different content page B will likely be ignored. Google prefers to index both or ignore the canonical and choose for itself. The reverse case also exists: a well-placed canonical on nearly duplicate pages will generally be respected, especially if the internal linking and technical signals converge.<\/p>

When does this rule not really apply?<\/h3>

If your site is small, young, and you rigorously control your internal linking<\/strong> and backlinks<\/strong>, the canonical is likely to be respected without issues. It’s in complex environments that problems arise: multi-country sites with hreflang, sites undergoing partial redesign where old and new URLs coexist, or UGC platforms where users generate chaotic internal links.<\/p>

The case of limited crawl budget<\/strong> is also a factor. If Google only crawls a fraction of your site on each visit, it may never encounter all the necessary signals to consolidate the canonical. You then end up with variants indexed "by default", due to insufficient signals to merge them. In this case, forcing with a noindex may sometimes become the only feasible solution, even if it means sacrificing PageRank consolidation.<\/p>

Warning:<\/strong> Never combine canonical and noindex on the same page — it’s contradictory. If you want to block the indexing of a variant, use noindex alone. If you want to consolidate the signal, use canonical alone. Mixing the two confuses the directives, and Google may ignore both.<\/div>

Practical impact and recommendations

What actions should be taken to avoid indexing issues?<\/h3>

Start with a canonical audit<\/strong>: crawl your site with Screaming Frog or Oncrawl, export all pages with a canonical tag, and check that they point correctly to the desired version. Then cross-reference with Search Console data to spot indexed pages that do not match your canonicals. If you discover indexed variants despite the canonical, dig into the backlinks and internal linking pointing to those variants — that’s often where the problem lies.<\/p>

Next, harmonize your internal linking<\/strong>. If you canonicalize from page A to B, all your internal links should point to B, not A. This is a common mistake on e-commerce sites: the canonical points to the product page without parameters, but the navigation links point to the sorted version. Google receives a contradictory signal and may opt to index the wrong version. Automate this check in your CMS or page template.<\/p>

What mistakes should be absolutely avoided?<\/h3>

Never chain canonicals<\/strong> (A → B → C). Google generally follows only one level of canonical. If you chain multiple, it may choose to ignore all of them and select the version to index itself. Also, ensure that you never point a canonical to a 301 or 404 page<\/strong> — this happens more often than one might think after a redesign or product removal.<\/p>

Another classic pitfall: canonicalizing to a page blocked in robots.txt<\/strong>. Google can't validate that the target of the canonical exists and is relevant, so it might ignore the signal. The same goes for pages in noindex<\/strong>: canonicalizing to a page in noindex is contradictory — Google cannot consolidate the signal to a page it shouldn’t index. In this case, it will likely index the source page or ignore both directives.<\/p>

How can I check if my site is properly configured?<\/h3>

Use Search Console<\/strong> to identify indexed pages that shouldn’t be. The "Pages" report shows excluded pages with the note "Duplicate, alternative page with appropriate canonical tag" — that’s a good sign. If you see "Indexed, not selected as canonical," it means that Google has indexed a different variant than the one you indicated. Investigate why: backlinks to the variant, contradictory internal linking, or conflicting hreflang signals.<\/p>

Set up an automatic monitoring<\/strong> of your canonicals. A script that regularly crawls your site and alerts you if an indexed page does not match its declared canonical. On high-volume sites, this is the only way to detect deviations before they impact traffic. Also, integrate a server log check to see which variants Google crawls the most — if it’s not your canonicals, you have a signal problem.<\/p>

  • Crawl the site and check the consistency of canonicals with the pages actually indexed in Search Console.<\/li>
  • Audit the internal linking to ensure that all links point to the canonical versions, not the variants.<\/li>
  • Check that no canonical points to a page in 301, 404, noindex, or blocked in robots.txt.<\/li>
  • Cross-reference external backlinks with indexed variants to detect conflicting signals.<\/li>
  • Implement automatic monitoring of canonicals and indexed pages to detect deviations.<\/li>
  • Consider the noindex tag on variants you absolutely do not want indexed, accepting the loss of PageRank consolidation.<\/li><\/ul>
    The canonical remains a powerful consolidation tool, but it guarantees nothing by itself. On complex sites with a significant backlink history or difficult-to-manage internal linking, managing signals can quickly become a technical headache. If you notice persistent variants despite your efforts, or if you want to secure your canonicalization strategy on a large site, the support of a specialized SEO agency can save you valuable time and avoid costly mistakes in crawl budget and signal dilution.<\/div>

❓ Frequently Asked Questions

Est-ce que le canonical transmet le PageRank contrairement au noindex ?
Oui, le canonical consolide le PageRank vers la page cible, tandis que le noindex bloque l'indexation ET le passage de PageRank. C'est la différence fondamentale entre les deux.
Combien de temps faut-il à Google pour respecter un nouveau canonical ?
Ça dépend du crawl budget et de la fréquence de passage de Googlebot. Sur un site à forte autorité, quelques jours à semaines. Sur un site lent à crawler, ça peut prendre plusieurs mois.
Peut-on canonicaliser une page vers un autre domaine ?
Oui, le canonical cross-domain est techniquement possible, mais Google peut l'ignorer si les deux domaines n'ont aucun lien éditorial évident. C'est rarement une bonne idée en pratique.
Que faire si Google indexe la mauvaise version malgré le canonical ?
Vérifie ton maillage interne et les backlinks vers la variante indexée. Si les signaux contradictoires sont trop forts, envisage un noindex sur la variante ou une 301 si elle n'a pas de raison d'exister.
Le canonical self-référencé sur chaque page est-il vraiment utile ?
Oui, il évite que des paramètres d'URL accidentels (utm, session ID) créent des variantes indexables. C'est une bonne pratique défensive, surtout sur des CMS qui génèrent des URL dynamiques.

💬 Comments (0)

Be the first to comment.

2000 characters remaining
🔔

Get real-time analysis of the latest Google SEO declarations

Be the first to know every time a new official Google statement drops — with full expert analysis.

No spam. Unsubscribe in one click.