LLMs and search work better when each language has its own clean URLs, sitemap, and clear alternates. Localize your answer hubs, wire hreflang accurately, and expose per‑locale discovery hints.
Architecture patterns
- One domain, language folders: example.com/en/, /es/
- Per‑locale subdomains: en.example.com, es.example.com (ensure separate sitemaps)
- Keep slugs human and stable per locale; avoid mixed‑language pages
hreflang essentials
- Declare reciprocal hreflang for every language/region pair
- Include an x‑default for fallback
- Keep canonical self‑referential
Example link tags (EN/ES):
<link rel="alternate" hreflang="en" href="https://example.com/en/faq" />
<link rel="alternate" hreflang="es" href="https://example.com/es/faq" />
<link rel="alternate" hreflang="x-default" href="https://example.com/faq" />
Per‑locale sitemaps and llms.txt
- Generate sitemap per locale/host; reference in each host’s robots.txt
- Optionally publish /llms.txt per host or directory, listing that locale’s hubs and txt/md assets
Content tips for LLMs
- Translate intent, not just words; adapt examples, units, proof
- Keep chunk sizes similar; maintain anchor names in‑language
- Reuse structure so models can map sections across locales
FAQs
- Do I need country and language codes (en-US vs en-GB)? If content differs meaningfully (currency, legal), yes.
- Can I use automatic translation? Use human review for high‑signal hubs; poor translations reduce trust and visibility.
- Is a single global llms.txt enough? It can help, but locale‑scoped hints improve discoverability of the correct language.
Sources
- Google Search Central: localized versions (hreflang)
- answer.cloud llms.txt guide
- answer.cloud AEO overview for foundations