AI tools now churn out polished documentation, but without clear source attribution, developers struggle to distinguish fact from inference. In a recent example, an API spec listed REST architecture, bearer tokens, rate limits, and RFC 7807 errors—none of which were flagged as user‑provided or AI‑generated. This opacity erodes trust today.

To restore clarity, the author proposes tagging every statement with its origin: [explicit] for user input, [inferred] for derived content, [assumed] for placeholders, and [general] for common knowledge. Inferred claims must include a verbatim quotation from the source, turning hidden assumptions into visible, verifiable evidence for future documentation and review.

Yet, terminology drift and multi‑user environments can still blur meaning. The solution calls for shared glossaries and versioned terminology, especially when new projects integrate AI from day one. Legacy systems should adopt the protocol only when modified, ensuring that source tags and consistent language keep documentation trustworthy for future teams.