Why does Google overlook your canonical on multi-country sites?

Why does Google group supposedly distinct URLs?

When a French e-commerce site publishes exactly the same catalog in French for .fr, .be, and .ch, Google considers this as duplication. It doesn't matter if you have correctly implemented hreflang and self-referential canonical tags on each version. The algorithms detect the identity of the content and operate clustering to avoid wasting crawl budget.

The result? Google chooses one single canonical URL from your country versions, often the one it deems most relevant according to opaque criteria: domain popularity, geographical signals, indexing history. Other versions end up as duplicates ignored, even if they target distinct markets.

Does hreflang become useless in this specific case?

No, but its role then becomes limited to display selection in the SERPs based on the user's geolocation. If Google has grouped .fr, .be, and .ch under a single canonical, hreflang can still direct a Belgian user to .be rather than .fr — as long as that version is actually indexed.

The trap: Search Console may show .be as “Alternate page with proper canonical tag”, which can create the illusion that everything is functioning properly. But if the content is identical, Google does not guarantee separate indexing. You end up with only one version indexed, and the others are in canonical limbo.

What does “significantly different” mean in Google's eyes?

Mueller provides no numeric threshold — as is typical. In practice, a simple change in currency, legal mentions, or footer is not enough. There must be substantial variations: localized product descriptions, editorial content unique to each market, different pricing structures with contextual explanations.

Concrete example? If you sell sneakers, don’t just change € to CHF. Write product descriptions tailored to Swiss preferences, add local guides (“Delivery in Switzerland: timelines and fees”), create region-specific editorial sections. Google needs to perceive a distinct editorial intent, not a copy-paste job with monetary variables.

• Algorithmic clustering: Google merges URLs with identical content to save on crawl, regardless of your hreflang/canonical structure
• Opaque selection: the canonical version chosen by Google does not necessarily correspond to your declared preference
• Conditional indexing: the other versions remain technically “known” but are not indexed separately
• Substantial differentiation required: cosmetic changes are insufficient; real editorial variations between markets are needed
• Hreflang retains a role: it can guide geolocated display, but does not enforce indexing if the content is duplicated

Is this statement consistent with real-world observations?

Absolutely, and it’s even a relief that Google finally acknowledges it publicly. For years, we have seen multi-country sites with perfect hreflang mysteriously lose national versions from the index. Search Console showed “Alternate page”, everything seemed technically correct, but .be or .ch URLs never ranked. The issue was not implementation; it was duplicated content.

What still shocks some practitioners: hreflang was marketed as the anti-duplicate solution for international sites. Mueller reminds us that no — hreflang indicates linguistic alternatives; it does not magically cleanse duplication. If the content is identical, Google applies its clustering logic, period. The tags then become simply display suggestions, not indexing orders.

What nuances should be added to this rule?

First point: “exactly the same content” remains vague. [To be verified] but we can assume that Google uses some form of internal similarity calculation, probably a derivative of shingling or fingerprinting. A textual duplication rate over 85-90% likely triggers the clustering. But again, no official figures, just empirical deductions.

Second nuance: the statement explicitly concerns “same content in the same language.” If you have a .fr in French and a .ca in English, the problem obviously doesn’t arise. But what about a .fr and .ca both in French, with Quebec orthographic variations? Totally gray area. My intuition: if the differences are limited to “colour” vs “color” or a few scattered Quebecisms, Google is likely to cluster anyway.

In what cases does this rule not fully apply?

If your site has a very strong domain-level authority and a long indexing history on each ccTLD, Google may tolerate more duplication without aggressively merging. I’ve seen big e-commerce players maintain several indexed country versions despite 95% identical content — probably because each domain generates significant traffic and local backlinks.

Another exception: sites with strong geographical signals outside content. Local hosting, backlinks from media in the target country, Google Business profiles by market — all of this can weigh on Google’s decision. But be careful, don’t rely solely on this: it’s support, not a solution. Differentiated content remains the only reliable strategy.

What should you concretely do to avoid this clustering?

First, honestly audit the degree of differentiation between your country versions. Open two URLs side-by-side (.fr vs .be for instance) and compare line by line. If only the header, footer, and a few legal mentions change, you’re in the red zone. You need to rewrite strategic content: product sheets, categories, service pages.

Next, create editorial content unique to each market. A blog with localized articles, buying guides tailored to local regulations, specific FAQs. The goal: that Google scans the page and detects a unique textual footprint. An identical template engine with replaced variables is no longer enough — you need tailored writing.

How to check that Google is correctly indexing each version separately?

In Search Console, connect each property (fr, be, ch) and compare the metrics of effective indexing. If .be shows 500 submitted URLs but only 50 indexed, with the other 450 as “Duplicate, Google chose different canonical”, you have proof of clustering. Cross-reference with site:votresite.be in Google: if the results mainly show .fr URLs, it means Google has made a decision.

Also, use the URL inspection tool on a few key pages of each version. Check the section “Google-selected canonical”: if it consistently points to .fr when inspecting .be, the diagnosis is clear. Be careful, this tool only shows the decision for the tested URL, not an overview — but repeating this on 10-15 strategic URLs will give you a reliable trend.

What mistakes should you absolutely avoid in this setup?

Do not cannibalize your own efforts by declaring a cross-canonical “for safety”. If .be points to .fr as canonical, you validate the clustering yourself — Google doesn’t even need to decide. Each version must have a strict self-referencing canonical, and hreflang must link the versions together without hierarchy.

Avoid the trap of automatically generated “spin content”. Just replacing a few words with synonyms via a script will not create sufficient differentiation — Google has detected these patterns for years. Authentic writing work is needed, not algorithmic camouflage. Finally, do not overlook local on-site signals: mentions of physical addresses, national phone numbers, native currencies displayed everywhere (not just a JS selector), all reinforce the perception of distinct pages.

• Differentiate at least 30-40% of the textual content between each country version (descriptions, editorial, FAQ)
• Create unique content sections for each market: local guides, regional case studies, customer testimonials from the country
• Check in Search Console that each property shows an indexing rate >80% of the submitted URLs
• Use the URL inspection tool to validate that Google respects your self-referencing canonicals
• Maintain strong geographical signals: local hosting if possible, backlinks from local media, LocalBusiness schema with addresses
• Monthly monitor indexing metrics to detect any gradual clustering