How does Google really weigh canonicalization signals?

Why does Google refer to 'weights' for canonicalization signals?

Canonicalization relies on a dozen signals that Google uses to determine which version of a page to display in search results. Among these signals are: the rel="canonical" tag, 301 redirects, internal linking, XML sitemaps, external backlinks, the URL displayed in the content, and HTTPS vs HTTP protocol.

Each signal has a relative weight—a variable importance depending on the context. Before machine learning, engineers manually set these weights. The problem? Boosting the weight of the canonical tag could inadvertently decrease the weight of redirects or the sitemap, leading to consequences that were impossible to predict.

What changed with the introduction of machine learning?

Google has delegated the task of continuously calculating and adjusting these weights to machine learning algorithms. The system analyzes millions of real-world cases to understand which signal proves to be the most reliable in a given context: an e-commerce site with 50,000 products is treated differently than a 200-page WordPress blog.

This automation makes the process opaque and unpredictable for SEOs. You could have a perfectly configured canonical tag and still see Google prefer a URL found in your sitemap—without any documentation explaining why in your specific case.

Does this challenge canonicalization best practices?

No, but it does put them into perspective. The best practices still apply: clean canonical tags, consistent redirects, up-to-date sitemaps, and unified internal linking. However, one should no longer expect absolute guarantees. Google provides levers, but it ultimately decides.

Gary Illyes' statement confirms what many observe in the field: sometimes, Google ignores your directives for no apparent reason. This isn't a bug—it's machine learning deciding that another signal deserves more trust in that context.

• Machine learning automatically weighs a dozen canonicalization signals
• Manual adjustment created unpredictable side effects between signals
• No single signal guarantees 100% the choice of the canonical URL
• Google judges based on context—site, theme, history, overall consistency
• Best practices remain relevant, but their effectiveness is no longer deterministic

Is this statement consistent with field observations?

Absolutely. We regularly see cases where Google ignores an explicit canonical tag to prefer a URL it detected via the sitemap or internal linking. Or the reverse: a poorly configured sitemap pointing to HTTP URLs when the site has migrated to HTTPS, and Google still chooses the HTTPS version thanks to backlinks.

What was frustrating was the lack of an official explanation. Now we know: there is no fixed hierarchy among signals. Machine learning judges in real-time, and its criteria are not documented—probably because they vary according to thousands of variables.

What nuances should we consider?

Gary Illyes does not specify which signals are included in the model, nor the exact number. We know that the canonical tag, redirects, sitemap, internal linking, and backlinks are part of it. But what about the URL displayed in the content? Hreflang? AMP annotations? [To be verified]—there is no official list.

Another point: Google does not state whether the machine learning model is unique for all sites or whether it adapts by sector, size, or type of CMS. Is a Shopify site with 100,000 products treated by the same rules as a 500-page WordPress? Probably not, but we’re navigating in the dark.

In which cases does this logic pose a problem?

When managing a highly duplicate site—e-commerce with filters, multilingual settings, SaaS platform with parameterized URLs—you need precise control. If Google decides that a signal you have not prioritized deserves more weight, you end up with non-canonical indexed URLs, duplicate content in SERP, and wasted crawl budget.

Machine learning optimizes for the average, not for your specific use case. If your site diverges from the norm—atypical architecture, custom CMS, complex business logic—you risk incoherent canonicalization decisions that go against your strategy.

What should you do concretely to maximize your chances?

Since you can no longer control which signal will weigh the most, the only viable strategy is absolute coherence across all signals. If your canonical tag points to URL A, your sitemap should list A, your internal linking should point to A, your redirects should lead to A, and your backlinks should ideally point to A.

Practically, this entails a regular audit: check that your CMS isn’t generating conflicting canonicals, ensure your XML sitemap doesn’t contain HTTP URLs if you’re on HTTPS, and that your internal linking does not mix www and non-www. Every inconsistency gives Google a reason to ignore your preferences.

What mistakes should you absolutely avoid?

Never allow conflicting signals to coexist. A classic example: a canonical tag pointing to URL A, but an XML sitemap listing URL B. Google will arbitrate, and you won’t know in advance who will prevail. Another mistake: chain redirects (A → B → C)—Google may decide that C is canonical when you intended B.

Avoid also multiplying URL parameters that are not properly managed. If you have filters, tracking, pagination, you must either explicitly canonicalize them, block them in robots.txt, or declare them as parameters in Search Console. Leaving it to Google to guess puts you at risk of errors.

How can you check that Google respects your intentions?

Use the Search Console: the "Coverage" tab and "URL Inspection" show you which URL Google has chosen as canonical for each page. Compare with your directives. If you notice discrepancies, dig deeper: which signal did Google favor? Sitemap? Linking? Backlinks?

Avoid also monitoring the server logs: if Googlebot heavily crawls URLs you've canonicalized to another, it's a signal that it did not heed your directive. Finally, a crawl using Screaming Frog or Oncrawl allows you to cross-reference all your canonical tags, redirects, and sitemap to detect inconsistencies before Google exploits them.

• Audit all canonicalization signals to ensure their absolute coherence
• Check that canonical tags, sitemaps, internal links, and redirects point to the same URL
• Use Search Console to compare the canonical URLs chosen by Google vs your directives
• Regularly crawl your site to detect contradictions before Google arbitrates
• Avoid chain redirects and undeclared URL parameters
• Monitor logs to spot heavily crawled non-canonical URLs