AI-Assisted Translation Policy
AI-generated translation may be used as a draft stage, but review, terminology control, and transparency remain required.
Rules
- AI output is a draft, not final truth.
- Sensitive wording needs explicit human review.
- Terminology should be checked against project guidance.
- Projects should disclose machine-generated status where useful.
Review expectations
Even when the overall meaning is preserved, AI-generated translations may require native-speaker review to correct subtle issues in grammar, modality, terminology, and domain-specific register. These issues are often difficult to detect through automated quality metrics alone.
Human review records should capture:
- original text
- corrected text
- error category
- root-cause hypothesis
- reviewer rationale
See ../style-guide/localization/ai-translation-review-records.md for the standard review-record format and case-study examples.
Let Books context
Let Books treats AI as a productivity tool and accessibility enabler, not a replacement for human linguistic review.