Multilingual Discovery for English-Only Sites¶

Making an English-only static site (MkDocs, Hugo, Jekyll on GitHub Pages) discoverable in multiple languages without full translation. The strategy: translate hub/landing pages per language, keep articles in English, use metadata and structured data to signal multilingual availability.

Key Facts¶

• Translated sites get 24% more total citations and 327% visibility boost cross-language in AI search
• ChatGPT uses English sources in 43% of non-English research steps; 78% include supplementary English search
• Google AI Overviews auto-translate English content when local sources are sparse (7.9% for Polish queries)
• Full translation is unnecessary - hub pages + proper metadata capture most of the discovery benefit
• 67% of hreflang implementations have errors - keep it simple, don't over-engineer

Strategy: Hub Pages + English Articles¶

docs/
  index.md              # English hub (default)
  index.ru.md           # Russian hub page (translated)
  index.zh.md           # Chinese hub page (translated)
  kafka/
    consumer-groups.md  # English article (stays English)
    index.md            # English domain index
    index.ru.md         # Russian domain index (translated)

Hub pages are 10-20 per language (domain indexes + main landing). Articles stay English. This captures discovery benefit at ~5% of full translation cost.

mkdocs-static-i18n Plugin¶

# mkdocs.yml
plugins:
  - i18n:
      default_language: en
      languages:
        - locale: en
          name: English
          default: true
        - locale: ru
          name: Russian
          build: true
        - locale: zh
          name: Chinese
          build: true
      fallback_to_default: true  # untranslated pages fall back to English

Suffix-based (index.ru.md) vs folder-based (/ru/index.md):

With Material theme, the language selector and hreflang tags generate automatically.

Hreflang Tags¶

Only valid for pages that actually have translations. Must be bidirectional:

<!-- On English hub page -->
<link rel="alternate" hreflang="en" href="https://site.com/" />
<link rel="alternate" hreflang="ru" href="https://site.com/ru/" />
<link rel="alternate" hreflang="x-default" href="https://site.com/" />

<!-- On Russian hub page -->
<link rel="alternate" hreflang="en" href="https://site.com/" />
<link rel="alternate" hreflang="ru" href="https://site.com/ru/" />
<link rel="alternate" hreflang="x-default" href="https://site.com/" />

Do NOT add hreflang to English-only article pages. Hreflang requires a corresponding page in the target language. Pointing to the English version from itself is meaningless.

Engine support:

Multilingual llms.txt¶

No formal spec for multilingual llms.txt yet. Emerging practice:

/llms.txt           # English (default), serves as index
/ru/llms.txt        # Russian language version
/zh/llms.txt        # Chinese language version

# Knowledge Base (Russian Index)

> Техническая база знаний: Kafka, Python, Rust, SQL, DevOps

## Основные темы
- [Kafka Consumer Groups](https://site.com/kafka/consumer-groups/): Партиции, оффсеты, ребалансировка
- [Python Async](https://site.com/python/async/): asyncio, aiohttp, паттерны

## О базе знаний
Статьи на английском языке. Эта страница помогает найти нужный контент.

AI bots DO request subdirectory llms.txt files. No confirmed AI company officially reads llms.txt at crawl time, but 844K+ sites have adopted it.

Schema.org for Multilingual Sites¶

{
  "@context": "https://schema.org",
  "@type": "WebSite",
  "url": "https://site.com",
  "inLanguage": "en",
  "availableLanguage": ["en", "ru", "zh"],
  "name": "Knowledge Base"
}

For translated hub pages, link to originals:

{
  "@context": "https://schema.org",
  "@type": "WebPage",
  "url": "https://site.com/ru/",
  "inLanguage": "ru",
  "translationOfWork": {
    "@type": "WebPage",
    "url": "https://site.com/"
  }
}

Critical: partial schema = 18% citation penalty. Only add translationOfWork / workTranslation if both pages exist and are properly linked. availableLanguage on the WebSite schema is safe to add always.

Search Engine Specifics¶

Yandex¶

• Supports hreflang (HTML head only)
• IndexNow for fast indexing of new pages
• Russian meta description required on Russian pages
• Register in Yandex Webmaster (free, ~1 hour setup)
• Separate sitemap with Russian pages helps

Baidu¶

• Does NOT support hreflang
• Uses <meta name="language" content="zh"> tag
• Strong bias toward Chinese-hosted content
• 1-3 month indexing delay for non-Chinese sites
• Unreliable for GitHub Pages - skip unless Chinese market is primary

AI Search Cross-Language Behavior¶

• ChatGPT: 43% of non-English research steps use English sources, 78% supplement with English
• Perplexity: returns English URLs even when answering in other languages
• Google AI Overviews: 96% language-matched citations normally, but auto-translates English when local content is sparse

English content has a structural advantage in AI search - it's the fallback for all languages.

Google Mixed-Language Policy¶

• No penalty for having some pages translated and others not, IF each page is one language
• Warns against translating only navigation/boilerplate
• Hub pages need substantial translated text, not just menu items
• Uses visible content (not HTML lang attribute) for language detection
• Each URL must be one consistent language

Cost-Benefit Ranking¶

Gotchas¶

• Machine-translated hub pages get flagged - Google flags machine-translated content without human review as auto-generated spam. Hub pages must be reviewed by native speakers before publishing. Use AI for first draft, human for final review
• Hreflang without matching pages breaks silently - adding hreflang="ru" pointing to a page that's actually in English confuses search engines. Google ignores it; Yandex may penalize. Only add hreflang when a real translated page exists at that URL
• inLanguage must match actual content - setting inLanguage: "ru" on an English article because it has a Russian meta description is incorrect. The property refers to the primary content language, not metadata language

See Also¶

• llm discoverability ai search - AI search optimization fundamentals
• regional seo - regional search engine optimization
• robots txt sitemaps indexation - crawl control and sitemap configuration
• technical content seo strategy - content architecture for discovery