Check the claims, not just the tone
A confident tone is not evidence. Highlight factual claims, names, dates, calculations, or recommendations that need checking against a reliable source or your own expertise.
If a claim matters to the decision, make its source visible before you rely on it.
Check whether the result fits the situation
A technically correct answer can still be unsuitable for the audience, task, or organisation. Read it for missing context, inappropriate assumptions, and details that should not be included.
This is also the point to remove confidential information and make sure the next reader understands what is draft material and what has been verified.
- Accuracy: Can the important claims be supported?
- Completeness: Is any needed context missing?
- Fit: Does the result suit the audience and purpose?
- Care: Does it expose information or assumptions that need attention?
Leave a visible human decision
End the review by deciding what you will change, verify, or ask next. That note can be as simple as a source to check or a sentence to rewrite.
A visible decision point keeps the workflow honest and makes the next use of AI more intentional.