How does Google choose the canonical URL among more than 20 signals?

What is a duplicate cluster and why does Google need to choose?

When several URLs display identical or very similar content, Google groups them into a duplicate cluster. This phenomenon occurs more often than one might think: HTTP/HTTPS variants, with or without www, tracking parameters, paginated versions, facet filters, syndicated content.

Instead of indexing each variant, the engine selects only one as the canonical URL — the one it judges most relevant for its users. The other URLs in the cluster remain known, sometimes crawled, but do not participate in ranking. This is an economic arbitration for Google and a consolidation of signals for SEO.

What are the 20+ signals mentioned in concrete terms?

Gary Illyes cites five explicit examples: content, PageRank, HTTPS, sitemaps, and redirects. The rest remains vague — deliberately. We can infer that Google also looks at internal consistency (massive internal links pointing to one variant), content freshness, hreflang tags, URL structure, Core Web Vitals, and the crawl volume allocated to each variant.

The order of priority is never revealed. One signal can weigh heavily in one context and become negligible in another. [To be verified] Google claims to use over 20 signals, but it's impossible to list them all or to know their relative weights — total opacity on internal scoring.

Why is the canonical tag just one signal among many?

Many SEOs still believe that placing a canonical tag is enough to impose their choice. This is a mistake. Google treats it as a suggestion, never as a directive. If other signals contradict this tag — for example, an alternative variant accumulates more PageRank or receives more external links — Google can ignore the tag.

This is consistent with the engine's logic: it never fully delegates the decision to a webmaster, who might misconfigure or attempt to manipulate. Canonicalization remains an algorithmic process, where every clue counts but none decides alone.

• Google groups duplicate content into clusters and chooses only one canonical URL per cluster
• Over 20 different signals are weighed — content, PageRank, HTTPS, sitemaps, and redirects explicitly mentioned
• The canonical tag is one signal among others, never an absolute directive
• The order of priority and exact scoring remain opaque, varying by context
• Google always keeps final control: it is impossible to force its choice 100%

Is this statement consistent with real-world observations?

Yes, largely. It is regularly observed that Google ignores poorly placed or contradictory canonical tags. A site that declares page A as canonical but redirects all its link juice to page B will often see B indexed. The same is true for sites that place canonical tags on active pages but neglect their XML sitemap — Google then prioritizes the version present in the sitemap.

The weight of PageRank in the canonical selection is empirically confirmed: a variant that accumulates quality backlinks often gets chosen, even if it is not the one the webmaster wanted. The effect of HTTPS also plays a role — since 2014, Google systematically favors the secure version when one exists.

What nuances should be added to this statement?

First point: [To be verified] the figure "more than 20 signals" remains unverified. It is impossible to know if Google uses 22, 50, or 150. This is a deliberately vague communication to discourage any attempt to game the system.

Second nuance: not all signals carry the same weight. HTTPS and 301/302 redirects are strong signals, almost imperative. A well-configured 301 redirect almost always overrides other clues. Conversely, a simple internal link to a variant weighs little if everything else points elsewhere.

Third point: consistency matters more than any isolated signal. A site that sends contradictory messages — canonical to A, sitemap with B, redirects to C — leaves Google to decide alone, often in unpredictable ways. The algorithm looks for consensus; if it does not find it, it applies its own rules.

In what cases can this rule cause issues?

Sites with facets suffer immensely. Each filter combination generates a unique URL — color, size, price. Google must then choose among hundreds of variants which one it will index. If the webmaster does not properly mark it up (canonical + robots), Google risks indexing low-value pages at the expense of the main categories.

Another critical case: content syndication. An article published on multiple partner sites creates multiple duplicates. Google will try to identify the original via freshness, links, and domain authority. But if all signals are equivalent, the engine may make a mistake and canonize towards a syndicator rather than the source.

What actions should be taken to master canonicalization?

First, audit the existing setup. Use Google Search Console, under Indexing > Pages, to identify URLs that Google has canonized differently than intended. Compare declared canonical URLs with those actually indexed. Any discrepancy indicates a conflict of signals.

Next, align your signals. If you want to enforce URL A as canonical, make sure that internal links massively point to A, that the XML sitemap lists only A, that 301 redirects converge to A, and that the canonical tag correctly points to A. Multi-signal consistency almost always overrides shaky configurations.

What mistakes should absolutely be avoided?

Never point a canonical tag to a URL that is 404 or 301 — Google ignores these tags and chooses arbitrarily. Never chain canonicals (A → B → C) — Google usually stops at the first jump or even ignores the entire chain.

Avoid also missing self-referencing canonicals. Every page should carry a canonical tag pointing to itself if it is the preferred version, or to the correct variant otherwise. The absence of a tag leaves Google to guess, with the risks that entails.

Another common pitfall: declaring a paginated page (page=2) canonical to page 1. Google does not like this — it considers that each page in a paginated series has its own distinct content. Preferably use rel=prev/next (obsolete but still observed) or let each page auto-canonicalize itself.

How can I check that my site is correctly configured?

Run a Screaming Frog or Oncrawl crawl while following the canonicals. Identify loops, chains, and broken canonicals. Ensure that 100% of strategic URLs carry a consistent canonical tag.

Then cross-check with Search Console: export the list of indexed URLs, compare it with your XML sitemap. Any discrepancy warrants investigation — either Google preferred another variant, or a contradictory signal is disrupting the algorithm. Dig into server logs to see which URLs Googlebot crawls the most: this is often an indirect indicator of the chosen canonical.

• Audit discrepancies between declared canonical and indexed URL using Search Console
• Align all signals: internal links, sitemap, redirects, canonical tag
• Avoid canonical chains, canonicals pointing to 404/301, and absence of tag on important pages
• Crawl the site to detect inconsistencies, loops, broken canonicals
• Cross-reference Search Console data, XML sitemap, and server logs to identify variants preferred by Google
• Prioritize strategic pages — categories, key product sheets, SEA landing pages