The fallback misconception: people expect en-us to cover en-ca, and it doesn't quite

The fallback misconception: people expect en-us to cover en-ca, and it doesn't quite

An analytical correction to a widely-held mental model: how Google actually picks among language and region variants, and the mistake that follows from getting it wrong.

The assumption we keep encountering: "I have en-us and en-gb; a Canadian English user will get whichever is closer." The model people imagine is a neat geographic nearest-neighbor. The real behavior is better described as: an exact language-region match wins; absent that, a same-language match (region-agnostic) is preferred; and an unmatched user falls to x-default.

The practical consequence — and the mistake:

— If you ship only region-locked variants (en-us, en-gb) with no plain en and no x-default, a Canadian or Australian English speaker has no clean target and the selection becomes less predictable
— Teams then "fix" this by adding ever more region locales (en-ca, en-au, en-ie…), multiplying near-duplicates instead of providing one general fallback

The better fix:

— Provide a region-agnostic en (and equivalently es, fr) as the catch-all for that language, reserving region codes for markets you genuinely differentiate
— Add x-default for users matching no language at all

Limitation and honesty: the exact selection algorithm is not fully documented and has nuance we can't observe directly — Google also factors signals beyond hreflang. What we can say from the documented behavior is that a plain-language fallback covers the long tail of regions far more efficiently than enumerating countries. Build the fallback; don't enumerate the world.