How does Google decide which version to index when you have duplicate content?

What does Google mean by "canonical version"?

The canonical version is the page that Google considers most representative of a set of similar or identical content. Contrary to what many believe, it's not necessarily the page you designated with your canonical tag.

Google collects multiple signals on each duplicate version: link popularity, internal linking consistency, URL structure, HTTPS protocol, technical performance. The algorithm aggregates this data to make its decision.

Why is this canonicalization so important?

Only the canonical page typically appears in search results. The other versions are consolidated: their signals (especially backlinks) are transferred to the chosen version, but they disappear from the visible index.

If Google chooses wrong — or if you haven't marked it correctly — you risk seeing a non-optimized version, a session URL, or even a test page appear in the SERPs.

What signals influence this decision?

• rel=canonical tag: strong signal but not binding
• 301 redirects: even more powerful signal than the tag
• Internal links: consistency of URLs pointed to
• XML Sitemap: declared URLs carry additional weight
• HTTPS vs HTTP: Google systematically prioritizes the secure version
• Technical quality: speed, accessibility, HTML structure
• External popularity: number and quality of backlinks to each version

Is this statement consistent with what we observe in the field?

Yes, but with important nuances. Gary Illyes confirms what practitioners observe: Google doesn't blindly follow your canonical directives. I've seen dozens of cases where the tag is ignored because other signals contradict it massively.

The problem? Google deliberately remains vague about the exact weighting of each signal. We know that 301 redirects carry weight, but what share do internal links represent versus external backlinks? Impossible to quantify. [To be verified] in each specific context.

In what cases doesn't this rule apply as expected?

When signals are contradictory. Imagine: you have a canonical tag pointing to /page-a, but 80% of your internal links point to /page-b, and /page-b receives 10 times more backlinks. Google will likely choose /page-b.

Another frequent case: multilingual sites with poor configuration. I've seen Google canonize an English version when the client wanted to push the French version, simply because technical signals (broken hreflang, inconsistent internal links) sent opposite messages.

Should you trust Google Search Console to identify issues?

Yes and no. GSC indicates which URL Google has canonicalized, but update delays are sometimes long. You fix an error, and GSC takes 3 weeks to reflect the change — meanwhile, it's impossible to know if Google has recrawled or not.

Let's be honest: GSC doesn't tell you why Google made this choice. It shows you the result, not the reasoning. To understand, you must cross-reference with server logs, crawl data, and internal linking analysis.

What concretely should you do to master canonicalization?

First, audit all duplicate URLs: session parameters, variants with/without slash, HTTP vs HTTPS, www vs non-www. Identify each duplication cluster.

Next, send coherent signals across all layers: canonical tags, 301 redirects if necessary, internal linking systematically pointing to the desired version, XML sitemap declaring only canonical URLs.

What mistakes must you absolutely avoid?

Never leave canonical chains: page A canonical to B, which itself is canonical to C. Google can follow, but it dilutes the signal and slows processing.

Also avoid crossed canonicals: page A points to B as canonical, and B points to A. It sounds absurd, but I've seen it on e-commerce sites with poorly managed filters.

Another classic pitfall: declaring a URL as canonical when it returns a 404 or 301. Google will ignore the directive and choose itself.

How do you verify your site is compliant?

• Crawl your entire site with Screaming Frog or Oncrawl to detect duplicates
• Verify that each page has ONE SINGLE canonical tag (self-referencing if it's the master version)
• Check in GSC the "Coverage" tab to spot "Discovered, currently not indexed" or "Excluded by a canonical tag" URLs
• Analyze server logs to see which versions Google crawls most
• Ensure all redirects are 301s (not 302s) to canonical URLs
• Verify internal linking consistency: no links to non-canonical versions
• Test hreflang tags if multilingual: they must point to canonical URLs