Picking an SSG for a Multi-Language Documentation Site

A multi-language documentation site multiplies every decision by the number of locales. Routing, content organization, fallback behavior, and build time all change character once you have four languages instead of one. The four major static generators — Astro, Hugo, Eleventy, and Jekyll — handle internationalization very differently, from Hugo's deep built-in support to Eleventy's convention-driven approach. This guide compares them on the dimensions that matter for translated docs, with measured numbers from a four-language, 1,000-source-page corpus. It sits under the SSG Framework Selection Matrix, within Choosing the Right Static Site Generator for Production.

Prerequisites

• A clear list of target locales and a fallback policy decision (what a reader sees when a page is not yet translated).
• A sense of your translation workflow — whether translators edit Markdown directly or content comes from a TMS export — since that shapes the content layout.
• hyperfine for build timing if you intend to compare candidates on your own corpus, plus each generator installed for a representative trial.

Routing Across Locales

Routing is where the gap is widest. Hugo treats languages as a first-class concept: declare them in hugo.toml and it generates /, /fr/, /de/ routes automatically with no plugin.

# hugo.toml
defaultContentLanguage = "en"
[languages.en]
  weight = 1
[languages.fr]
  weight = 2
[languages.de]
  weight = 3

Astro added native i18n routing in v4 — configure i18n in astro.config.mjs and it handles locale prefixes and getRelativeLocaleUrl helpers:

// astro.config.mjs
export default defineConfig({
  i18n: {
    defaultLocale: "en",
    locales: ["en", "fr", "de"],
    routing: { prefixDefaultLocale: false },
  },
});

Eleventy has no built-in locale router; the common approach is folder-per-language plus the community eleventy-plugin-i18n for string lookups, with permalinks driven by directory data. Jekyll relies on a plugin such as jekyll-multiple-languages-plugin or manual collections, which is the most hand-wired of the four.

Content Organization

For documentation at scale, folder-per-language is the layout that stays legible:

content/
  en/getting-started/index.md
  fr/getting-started/index.md
  de/getting-started/index.md

Hugo, Astro, and Eleventy all map this cleanly to per-locale routes. Hugo also supports filename suffixes (index.fr.md), which is tidy for small sites but noisy across thousands of pages. Jekyll typically models languages as collections, which works but couples your locale structure to Jekyll's collection conventions. Whichever generator you pick, mirror the directory tree exactly across locales so translators always know where a page lives — the same discipline that keeps large single-language docs maintainable, as discussed in Choosing Between Astro and Eleventy for Large Docs.

Fallback and hreflang

Two i18n-specific concerns separate a polished multi-language site from a broken one. Fallback: when a page is not yet translated, Hugo can serve the default-language version automatically; with Astro and Eleventy you decide explicitly whether to render the default page or hide the link, so readers never hit a 404. hreflang: each translated page should declare its alternates so the right locale is matched. Hugo emits these from its translation data; in Astro and Eleventy you build them from a small data structure listing each page's available locales:

{% raw %}{% for locale in page.locales %}
<link rel="alternate" hreflang="{{ locale.lang }}" href="{{ site.url }}{{ locale.url }}">
{% endfor %}{% endraw %}

Build Cost

Each locale is a full set of pages, so build time scales roughly linearly with total rendered pages. Timed with hyperfine --warmup 1 --runs 5 on the four-language corpus (1,000 source pages × 4 locales = 4,000 rendered pages):

The four-language numbers track the single-language ratios, confirming that i18n itself adds little overhead beyond the extra pages. Hugo's lead widens in absolute terms simply because it is the fastest per page. If build time at scale is your constraint, that ranking matches the broader picture in Hugo Build Times for Large Repositories, and the caching levers in Speeding Up Hugo Builds with Render Hooks and Caching apply per locale.

Putting It Together

For a large multi-language documentation site, the ranking is clear: Hugo if you want the deepest built-in i18n and the fastest builds; Astro if you want native i18n routing with a component model and can absorb the higher per-page build cost; Eleventy if your team prefers templates-and-Markdown and you accept wiring locale routing yourself; Jekyll only if you are already committed to it, since its i18n story is the most manual. Score these against your own weights using the SSG Framework Selection Matrix rather than picking on i18n alone.

Pitfalls & Rollback

• Inconsistent directory trees across locales: if fr/ and en/ diverge in structure, translators and fallback logic both break. Mirror the tree exactly.
• Missing fallback policy: without an explicit decision, untranslated pages 404. Pick fallback-to-default or hide-the-link and apply it uniformly.
• Forgetting hreflang: translated pages without alternate declarations leave locale matching to guesswork. Generate hreflang from a single source of locale data.
• Underestimating build time: four locales is four times the pages. Benchmark the full multiplied corpus, not a single language, before committing.
• Rollback: because all four generators consume a folder-per-language Markdown tree, a trial migration of one locale is cheap. Keep content generator-agnostic and you can build the same translated tree with another tool to compare real numbers.

Conclusion

Internationalization is the dimension where these four generators differ most. Hugo leads on built-in routing, fallback, hreflang, and raw build speed; Astro's v4 routing makes it a strong second; Eleventy and Jekyll handle multiple languages but ask for more manual wiring as locales grow. Decide your fallback policy, mirror the directory tree across locales, emit hreflang from one data source, and benchmark the full multiplied page count. Then weight i18n alongside your other criteria in the parent SSG Framework Selection Matrix.

FAQ

Which SSG has the strongest built-in internationalization?

Hugo has the most complete built-in i18n: native language configuration, automatic per-language routing, translation tables, and per-language sitemaps without plugins. Astro added solid native i18n routing in version 4. Eleventy and Jekyll handle multiple languages but lean more on conventions and add-ons.

How does build time scale with more languages?

Roughly linearly with total page count, since each locale is a full set of pages. In our four-language, 1,000-source-page test the 4,000 rendered pages built in 22 seconds with Hugo and 71 seconds with Astro, tracking the single-language ratio between the two.

Should each language live in its own folder or use filename suffixes?

Both work. Folder-per-language (content/en, content/fr) is the clearest for large docs and is the idiomatic layout in Hugo and Astro. Filename suffixes like page.fr.md are convenient for small sites but get noisy at scale.

How do I handle untranslated pages?

Decide on a fallback policy. Hugo can fall back to the default language for missing translations; with Astro or Eleventy you typically render the default-language page or hide the link. Define this explicitly so readers never hit a 404 for a page that exists in another language.

Do I need hreflang tags on a static multi-language site?

Yes, so each translated page declares its alternates. Hugo can emit these from its translation data; in Astro and Eleventy you generate them from a small data structure that lists each page's available locales.

Related

• Parent: SSG Framework Selection Matrix — the scored decision framework to weigh i18n alongside other criteria.
• Best SSG for Technical Writers Without Coding Experience — the authoring-first view that matters when translators edit Markdown.
• Choosing Between Astro and Eleventy for Large Docs — scaling discipline that compounds across locales.
• Hugo Build Times for Large Repositories — why Hugo's per-page speed leads the build-cost table.