Conflicting Hreflang and Canonical

What this conflict actually is

hreflang and rel canonical are both indexing signals, and they answer different questions. hreflang answers "which language and regional versions of this content exist, and who should see which one?" The canonical answers "of a set of duplicate or near-duplicate URLs, which single one should be indexed?"

They cooperate fine until someone points a canonical from one language version to another. The classic case: the French page at /fr/pricing/ declares hreflang alternates for English and French, but its canonical tag points at /en/pricing/. Now the page is simultaneously saying "I am the French alternate, index me for French users" and "I am a duplicate of the English page, do not index me at all."

Google's documentation and years of statements from John Mueller are consistent on how this resolves: hreflang annotations are processed on canonical, indexed URLs. A page that canonicalizes away is not the canonical, so its hreflang annotations get thrown out, and because that page was part of everyone else's annotation set, the whole cluster degrades. One bad canonical can quietly poison the hreflang for every locale.

This usually is not malice or even a decision anyone remembers making. It is an SEO plugin defaulting all canonicals to the "main" language, a dev copying the English template, or a CMS that treats translations as duplicates. I have seen a five-figure-page international site where every single non-English canonical pointed at English because of one Yoast setting. Nobody noticed for a year; Spain was being served English pages the whole time.

CONFLICT: canonical crosses the language boundary /en/pricing/ canonical: /en/pricing/ indexed /fr/pricing/ canonical: /en/pricing/ canonicalized away, hreflang dropped "index the English page instead of me" Result: French page drops out, French users get the English URL, the cluster is broken for every locale.

Why Google reacts the way it does

The mechanics matter because they explain the weird symptoms. Google builds hreflang clusters from annotations found on indexed canonical URLs. When your French page canonicalizes to English, three things happen in sequence:

First, Google eventually honors the canonical (it is also just a hint, but a strong one) and drops the French URL from the index in favor of the English one. Second, the hreflang annotations on the French page stop counting, since the page is no longer a canonical. Third, the annotations on your other pages that point at the French URL now target a non-canonical URL, which makes them invalid too. The English and German pages are pointing at a ghost.

The failure is gradual and partial, which makes it a miserable thing to debug. Some queries still show correct swapping for weeks because the cluster is reprocessed page by page as Google recrawls. Traffic in one country erodes over a quarter rather than falling off a cliff, and it is damn hard to spot in aggregate dashboards. Country-segmented reporting is what exposes it.

Scenario reference: which combinations work

How to detect it

Screaming Frog. Enable hreflang extraction, crawl, then open the hreflang tab and look at the "Non-Canonical Return Links" and "Canonical Errors" style filters (naming varies slightly by version). These list every annotation pointing at a URL whose canonical is elsewhere, and every page whose own canonical disagrees with its hreflang membership. Sitebulb flags the same pattern as a hint with the affected URL pairs.

Search Console. Check the Pages report for "Duplicate, Google chose different canonical than user" and "Alternate page with proper canonical tag" entries on URLs that should be independently indexed language versions. A translated page showing up there is the conflict in action. The URL Inspection tool shows the user-declared versus Google-selected canonical for any single URL, which is the fastest way to confirm a suspicion.

Country-level SERP checks. Run your key queries through a location-set VPN or a rank tracker with per-country tracking. Wrong-language URLs ranking in a market you have a local page for is the end-user symptom of this whole mess.

How to fix it, step by step

1. Export every URL in the multilingual set with its canonical and its hreflang annotations (Screaming Frog does this in one export).

2. Fix canonicals first: every language and regional version gets a self-referencing canonical. Hunt down the source, usually a plugin setting, a hardcoded template value, or a CMS translation module treating locales as duplicates.

3. Then align hreflang: every annotation must target the clean, 200-status, self-canonical URL of each version. No parameters, no redirects.

4. If two versions genuinely are identical and you only want one indexed, remove the duplicate from the hreflang cluster entirely and canonicalize it. A page is either an alternate or a duplicate, never both.

5. Resubmit sitemaps, then let Google recrawl the whole cluster before judging. Expect weeks, not days, for full reprocessing. Track recovery by country in the performance report.

FAQ