The standard advice for international SEO is subfolders.
site.com/de/ for German.site.com/fr/ for French.site.com/ja/ for Japanese.
This advice is wrong. Not slightly wrong. Fundamentally wrong. And following it is costing operators enormous amounts of traffic and revenue.
Here's why.
When Google crawls site.com/de/, it sees a German page on an English domain. The domain's authority, its backlink profile, its age and trust — all of that is anchored to the root domain's primary language and market. The German subfolder is essentially a guest in someone else's house.
When Google crawls de.site.com, it sees something different. A German website. With its own identity. Its own hreflang structure pointing outward to 68 other language variants. Its own canonical authority. Backed by the full domain authority of the root, but expressed as a genuine German web presence.
Google treats them differently. I've watched this happen in real time.
Within 24 hours of launching pl.pornfreetrial.com — the Polish subdomain of a site I rebuilt this week — we ranked #3 in Polish Google for our primary target term.
24 hours.
On a domain that had just been migrated from WordPress to a custom FastAPI stack.
That doesn't happen with subfolders. It happens because Google saw a real Polish website, with Polish copy, Polish currency (zł3.74 — "Praktycznie Za Darmo," basically free), proper hreflang tags pointing to 68 other genuine locale variants, and an IP detection layer that served Polish content to Polish visitors automatically.
That's the difference between translation and localization.
Translation is changing the words.
Localization is changing the experience.
The infrastructure I built through TreeChain does this across 69 languages from a single codebase. One offers.json. One FastAPI service. Wildcard DNS on Cloudflare pointing *.domain.com to a single server. IP detection routing visitors to their locale automatically.
The result: 69 genuine web presences, each with local currency, local copy that reads naturally (not machine-translated), and proper SEO structure — all deployed simultaneously on day one.
Most operators spend years building one market properly. This infrastructure builds all markets simultaneously.
The mistake everyone makes is thinking i18n is a translation problem. It's an infrastructure problem. Solve it at the infrastructure layer and translation becomes a consequence, not a project.
Subdomain theory isn't a trick. It's what correct looks like.