Subdomain vs Subdirectory for International SEO: A Decision Framework

Every international SEO discussion reaches the same argument within ten minutes, and it is usually argued from anecdote. The honest answer is that all three structures rank, the differences are smaller than the debate suggests, and the right choice depends on facts about your business rather than a general rule.

Should I use subdomains or subdirectories for multilingual sites?

Subdirectories for most businesses, because authority consolidates on one domain. Choose country-code domains when you need maximum local trust and have resources to build each one separately. Choose subdomains mainly when platform or infrastructure constraints force separate hosting.

The three structures, honestly compared

Subdirectories

Every market lives on the same domain, so links earned anywhere strengthen everywhere. Cheapest to run and simplest to manage. The weakness is genuine: they signal less local commitment, and in markets where buyers check for a local domain, that matters.

Country-code top-level domains

The strongest local trust signal and the clearest geotargeting. Also the most expensive, because each domain builds authority from zero. A new one can take a year to become competitive. Correct for businesses with real local operations and separate country teams.

Subdomains

Sit between the two and are usually chosen for infrastructure reasons: different platforms, different teams, different hosting per region. Authority sharing with the root domain exists but is weaker than with subdirectories.

The questions that decide it

1. Do you have local teams building local links? If yes, country domains become viable. If your link building is centralised, subdirectories keep that value in one place.
2. Is your product legally or commercially different per country? Genuinely separate businesses justify separate domains.
3. How strong is your existing domain? The stronger it is, the more you lose by fragmenting it.
4. Do buyers in your category expect a local domain? In some markets and sectors this is a real conversion factor, independent of ranking.
5. Can you maintain it? Five domains means five technical stacks, five sets of Search Console properties and five content calendars. Most teams overestimate this capacity.

Hreflang is where implementations actually fail

Structure choice rarely sinks a project. Hreflang usually does. The rules that catch people out:

• Every annotation must be reciprocal. If A points to B, B must point back to A, or both are ignored.
• Use self-referencing tags. Each page must include an annotation pointing at itself.
• Language code first, region optional. Use en for English generally, en-GB only when the content is genuinely specific to that region.
• Add x-default for your fallback page, usually the international or language-selection version.
• Point at the canonical URL, never at a page that canonicalises elsewhere. This combination silently voids the whole set.

The mistake underneath most failures

Publishing the same content in the same language across several country versions, differing only by currency, then wondering why one version outranks the rest. That is not internationalisation, it is duplication with a flag on it. Either genuinely localise, meaning different examples, pricing, regulations and phrasing, or serve one page and handle currency dynamically.

Translation is the minimum. Localisation is what actually ranks.

A free audit validates hreflang reciprocity and canonical conflicts across your locales.