rel=canonical vs hreflang: the conflict that silently kills clusters

rel=canonical vs hreflang: the conflict that silently kills clusters

Canonical tags and hreflang serve opposite purposes, and when they disagree, hreflang loses. We see this conflict in a striking share of broken clusters.

The core tension:

— rel=canonical says 'this other URL is the real one, index that instead'.
— hreflang says 'these URLs are equals, each for a different audience'.

When a localized page (example.com/fr/) carries a cross-language canonical pointing to example.com/en/, you have told Google the French page is a duplicate to be folded away — while simultaneously telling it via hreflang that French is a distinct equal. Google resolves this by trusting canonical and discarding the hreflang relationship.

What the data suggests:

— Every page in an hreflang set must self-canonicalize (canonical points to itself). This is the single most violated rule we audit.
— Cross-language canonicals are almost always a CMS misconfiguration, not intent.
— The damage is invisible in rankings dashboards; you only catch it by inspecting canonical and hreflang together, page by page.

Our tactic comparison: never use canonical to consolidate language variants. Use it only within a single language for genuine duplicates (parameters, print versions). Caveat — there are rare valid cross-locale canonicals, e.g. identical-content regions you deliberately consolidate, but treat them as exceptions requiring justification.