SEO translation: how to translate content that ranks in every language

Definitions

SEO translation vs SEO localization: what's the difference

SEO translation converts your content into another language while keeping it optimized for search: the target keyword, headings, metadata, and links survive the translation. SEO localization goes further and adapts the page to the market, with local search terms, currencies, units, examples, and cultural references. In practice the two overlap, and most sites need a bit of both: translate every page, then localize the pages that drive revenue.

• Translation: the same page, accurately rendered in a new language.
• Localization: the same intent, adapted to how a market searches and buys.
• Both fail if search engines cannot crawl and index the translated page.

Keywords

Research keywords in each language instead of translating them

A word-for-word keyword translation is often not what people search. German shoppers may use an English loanword, Spanish searchers in Mexico and Spain may use different terms, and the literal translation of your head term can have no search volume at all. Start from the English keyword's intent, then find the phrase local searchers use for it.

• Translate the intent, then validate the phrase in a local keyword tool for that country.
• Check what local competitors and the local search results use.
• Map one primary keyword to each translated page, the same as in your source language.
• Write the translated title, H1, and slug around that local keyword.

On-page

Translate everything search engines read, not just the body text

Many translated sites rank poorly because only the visible paragraphs were translated. Search engines also read the page title, meta description, URL, headings, image alt text, link text, structured data, and social tags. If those stay in English, the page sends a mixed-language signal and its search snippet shows the wrong language.

• Title tag and meta description, written for the local keyword.
• URL slug, for example /fr/tarifs instead of /fr/pricing, if your routing supports it.
• Headings, image alt text, button labels, and internal link text.
• Structured data, Open Graph, and the html lang attribute.

import { headers } from "next/headers";
import { altified } from "@/lib/altified";

export async function generateMetadata() {
  const locale = (await headers()).get("x-altified-locale") || "en";

  return altified.translatedMetadata({
    locale,
    path: "/pricing",
    origin: "https://example.com",
    locales: ["en", "fr", "de"],
    title: "Pricing",
    description: "Plans for your product.",
  });
}

Rendering

Serve translated pages as HTML that search engines can crawl

Search engines index what the server returns. JavaScript widgets that swap text after the page loads, and language switchers that change content without changing the URL, can leave crawlers with the source language or with one URL for every language. Each language needs its own URL that returns fully translated HTML on the first request.

• Give every language a stable URL: /fr/page, fr.example.com, or a country domain.
• Render the translation on the server, not after hydration.
• Do not redirect visitors or crawlers automatically by browser language.
• Translation proxies can work, but check that they serve complete, indexable pages under your own domain.

Contenuti dinamici

Translate dynamic and database content too

Product names, listings, help articles, reviews, and other records stored in a database are often the pages with the most search demand, and they are the ones translation projects usually miss. A translated template around English records still ranks as an English page. Translate content where it is produced, and keep translations in sync when records change.

• Translate records and API responses for the active locale, not only interface strings.
• Re-translate when the source record changes, so languages do not drift.
• Include translated record pages in your sitemap.

AI translation

AI translation and SEO: when machine-translated content ranks

Google does not penalize translation itself, but its spam policies name automatically translated text published without human review or curation as an example of scaled content abuse. Machine and AI translation can rank when the result is accurate, useful, and reviewed where it matters. The risk is publishing unreviewed output at scale: wrong product terms, awkward phrasing, and pages nobody in that market would trust.

• Use a glossary so brand names, product terms, and keywords translate consistently.
• Have a fluent reviewer check your highest-traffic and highest-revenue pages.
• Translate metadata and keywords deliberately rather than leaving them to the default output.

Signals

Connect translated pages with canonical, hreflang, and sitemaps

Once each language has its own crawlable page, tell search engines how the pages relate. Each translation needs a self-referencing canonical, a reciprocal hreflang set listing every language version, and an entry in your sitemap. Translated pages are not duplicate content when these signals are correct.

• Self-canonical on every language version.
• Reciprocal hreflang, including x-default where you have a default page.
• Every published language URL in the sitemap.
• Validate after each release.