AI-Translated Localization

When to invoke

Scope decisions:

Starting a new App and deciding which locales to support.
Choosing a string catalog format (.strings vs .xcstrings).
Planning the translation flow (manual / agency / AI agent).
User asks "are 7 locales too many", "how do translations enter git", "how to handle multi-locale App Store metadata".

Execution:

Adding new user-facing strings (UI keys, GC titles, ASC metadata, achievement descriptions) and need to fill out their non-source locales.
Refreshing strings whose source text changed after translation shipped (locales showing stringUnit.state: needs_review in xcstrings — not to be confused with extractionState: stale, which flags a key no longer referenced in code, not a changed source).
Auditing whether a release's xcstrings is complete (every key has every locale, no <TRANSLATE> placeholders shipping).