Better i18n/Help Center

How does AI translation work?

5 min readBeginner

Better i18n uses large language models (LLMs) to translate your content. Unlike simple machine translation, it understands context, follows your glossary, and adapts to your product's tone.

How to translate with AI

  1. Go to your project dashboard
  2. Select the target language (e.g., Turkish)
  3. Click "AI Translate"
  4. Choose scope:
    • All untranslated keys — fill in everything that's missing
    • Selected keys — translate specific keys
    • Re-translate — overwrite existing translations
  5. Review and confirm

What makes it context-aware?

The AI considers several signals when translating:

  • Key name — auth.login_button tells the AI this is a UI button
  • Context field — if you added context like "Shown on the login page header"
  • Glossary terms — your glossary enforces consistent terminology (e.g., always translate "workspace" as "çalışma alanı" in Turkish)
  • Surrounding keys — keys in the same namespace provide thematic context
  • Source value — the actual text to translate

Speed and limits

Volume Approximate time
1–50 keys A few seconds
50–500 keys Under a minute
500+ keys 1–3 minutes

AI translation runs in parallel batches for speed.

Quality tips

  1. Add context to ambiguous keys — "Save" could mean different things; "Save button in settings page" is clear
  2. Maintain your glossary — consistent terminology across all translations
  3. Review critical content — AI is great for speed, but review user-facing copy for tone
  4. Use the review queue — assign reviewers to verify AI translations before publishing

AI + human workflow

The recommended workflow combines AI speed with human quality:

  1. AI translates all keys (first pass)
  2. Reviewers check critical keys (review queue)
  3. You publish after review approval

You can assign team members with the Translator role to review AI translations. They'll see flagged keys in their review queue.

Next steps

Was this article helpful?