Subdirectory, Subdomain or Country Domain: Choosing a Multi-Language Structure
The multilingual decision is not an SEO question, it is an operational one. A framework based on who writes the content, plus the three ways subdirectories fail.
Every exporter building a multilingual store eventually faces the same question: subdirectory, subdomain, or separate domains. Most articles answer it with a table of pros and cons that ignores the only thing that really decides the outcome — how you will maintain the content.
Here is the decision framework we use, and the failure modes we see on stores that chose the wrong one.
The three options, stated plainly
| Structure | Example | What it optimises for |
|---|---|---|
| Subdirectory | example.com/de/ | Link equity concentration, simplest maintenance |
| Subdomain | de.example.com | Independent deployment, separate teams |
| Separate domains | example.de | Local trust, country-specific perception |
Google treats subdirectories and subdomains almost identically these days. It is not a ranking choice. It is an operational choice with SEO consequences.
What actually decides it
Who writes the content. If the same small team maintains every language, use subdirectories. One install, one plugin list, one deployment. Every extra install multiplies maintenance cost by the number of languages, and translation is already the expensive part.
If each market has its own team that deploys independently, subdomains or separate domains remove the coordination problem. That is the only strong argument for them.
Whether you need local trust. A .de domain carries a small trust advantage with German customers, mostly psychological. If your German market is significant and your brand is unknown there, a local domain can be worth the operational cost. For most exporters in their first three markets, it is not.
Whether the catalogue differs by market. If Germany sells a different product range from the US, that is a reason for separate structures. If it is the same catalogue with translated descriptions, subdirectories win clearly.
The cost nobody budgets for
Each additional structure multiplies four things:
- Plugin and theme updates — one per install, and they drift apart within six months
- Security surface — every extra WordPress install is another entry point
- Content consistency — the German site gets updated, the French one is forgotten
- Debugging — “the price is wrong in Spanish” now means checking which of six installs is wrong
On subdirectories, all four are single events. On six separate installs, they are six.
What breaks on subdirectories anyway
Subdirectories are the right default, but they fail in three predictable ways.
hreflang done halfway. The most common failure: translated pages output hreflang tags but the default-language page does not, or the tags are not reciprocal. When one page in the set omits them, search engines may discard the annotations for the whole group. Every page in the cluster must reference every other page including itself, plus an x-default.
Untranslated URLs. Slugs left in English while the page is in German create URLs that are hard to share and harder to remember. Localised slugs are better for users, but only if you commit to doing it consistently — mixing them creates the kind of mess where nobody can guess a URL.
Machine translation with no review. This is the biggest risk and the hardest to fix after the fact. A store with eight languages, all machine-translated, reads as low quality to both customers and search engines. Two well-written languages beat eight machine ones. If budget is limited, translate the pages that convert — home, category, product, checkout — and leave the blog in one language.
The switching cost is asymmetric
Moving from subdirectories to subdomains later is painful: every URL changes, every backlink needs a redirect, every ranking resets for a while.
Moving from subdomains to subdirectories is worse.
Which means the decision should be made once, early, with a clear head. The default answer — subdirectory, one install, hreflang correctly implemented, only the commercially important pages translated — is right for the large majority of exporters.
Where the money actually goes
Stores routinely spend on structure and not on translation quality. The structure is a one-time configuration. The translation is ongoing, and it is what customers read.
A single install with two excellent translations will outperform eight mediocre ones on every metric that matters, including how many languages Google bothers to index well. If you are planning a multilingual build, the question worth asking is not “how many languages can we launch with” but “how many can we keep genuinely good”.
If you want the technical setup reviewed — hreflang reciprocity, localised slugs, indexation of each language version — that falls under technical SEO, and it is worth doing before you add a ninth language.