SEO translation: how to rank in languages you don't speak

SEO translation is the practice of translating and adapting your pages, keywords, and metadata so they rank in search results in each target language, not just read correctly. globalize.now is AI-powered localization infrastructure that keeps the string side of that work, interface copy and metadata in locale files, in sync on every Git push. This guide covers how SEO translation differs from plain translation, the technical layer search engines require, and the workflow that holds up as your app keeps changing.

What is SEO translation?

SEO translation converts a page from one language to another while preserving, and re-targeting, its ability to rank: the keywords people actually search in that language, the metadata search engines read, and the technical signals that tell Google which version to serve.

Plain translation optimizes for accuracy. SEO translation optimizes for demand. The difference shows up in three places: the vocabulary your translated content uses, the title and description each translated page ships, and whether search engines can find and trust the translated version at all.

The prize is real. CSA Research surveyed 8,709 consumers across 29 countries and found 76% prefer buying with information in their own language, and 40% will not buy in other languages at all. Ranking in a market's language is how that preference finds you.

How is SEO translation different from regular translation?

A correct translation can still target a phrase nobody searches.

Search demand is expressed in each market's own vocabulary, and it rarely matches the dictionary translation. A car rental page targeting Spain wants alquiler de coches; the same page for Mexico wants renta de autos. Both are Spanish. A literal translation would pick one phrasing and silently forfeit the other market's search volume.

That is why SEO translation adds a keyword step: before translating a page, you check what the target market actually types for that intent, then translate toward those terms. The same applies to your metadata. A translated page shipping an English title and meta description loses the snippet before anyone reads the content.

Why don't translated pages rank?

Usually because search engines never see them as real, independent pages. Four failure modes cover most cases.

• Client-side translation. Runtime widgets translate the page in the browser after load. The served markup stays English, so the translated version has nothing for a crawler to index.
• No per-language URLs. If every language renders at the same URL, there is no stable address for a translated result to rank at. Each locale needs its own path, like /es/ or /de/.
• Missing or broken hreflang. hreflang annotations tell Google which language version to show which searcher. Get them wrong and Google may serve your Spanish page to English searchers, which craters click-through on both.
• Untranslated metadata. Body translated, title and description still English. The page may get impressions and still lose every click to a native-language snippet.

Note what is not on the list: translated content being treated as duplicate content. Properly annotated language versions are alternates, not duplicates. The risk with machine translation is quality, not duplication, so unreviewed raw output on money pages is the thing to avoid.

How do you translate a site for SEO, step by step?

The durable order is infrastructure first, then content, then signals.

1. Internationalize the codebase. Strings out of components and into locale files, per-locale routing, server-rendered translations. This is the i18n groundwork everything else stands on.
2. Check keywords per target market. For each page, find what that market searches for the same intent, and translate toward those terms rather than word for word.
3. Translate metadata as first-class strings. Titles, descriptions, Open Graph tags, and structured data belong in your locale files next to the UI copy, so no language ships an English snippet.
4. Wire the signals. hreflang annotations across all language versions, one x-default, and every locale URL in the sitemap.
5. Keep it in sync. Every release adds and changes strings. If translations lag, translated pages decay into mixed-language pages, and rankings decay with them.

What changes when an AI agent maintains your locale files?

Step five stops being a manual process, which is the step where most multilingual SEO quietly dies.

In an AI-built app the UI changes on every prompt, so the gap between shipped strings and translated strings reopens constantly. globalize.now closes it at the infrastructure layer: it extracts new and changed strings on every Git push, generates the translation keys, and returns updated locale files in a pull request. Because translated page titles and descriptions are strings in those same files, your metadata stays translated as the app evolves, with no extraction passes between releases.

That leaves the parts worth human judgment, market keyword choices and review of high-stakes pages, on top of locale files that are complete by default. The developer guide covers the file structure, and the vibe-coder setup shows the same flow from Cursor, Claude Code, or Lovable.