1. Name the editorial problem before asking for prose
Start with a diagnosis: unclear, unsupported, too strong, poorly connected, repetitive, or overly long. A named problem gives both the tool and the reviewer a boundary. “Improve this” is not a useful boundary because it invites changes to tone and intent.
Write the instruction as a testable request. For example: “Qualify the conclusion and keep the limitation in the same sentence.”
2. Mark what must stay stable
Identify the facts, commitments, terminology, and viewpoint that cannot move. In a research report that may be the study population and caveat. In a grant proposal it may be the funded objective. In a consulting memo it may be the decision logic.
- Facts and numbers
- Named evidence and attribution
- Scope, uncertainty, and qualifications
- Commitments, eligibility language, and legal terms
- The author's point of view
3. Compare meaning before style
Read the original and proposed sentence for meaning first. Ask whether the revision changes who is acting, how certain the claim is, what evidence supports it, or what decision follows. Only after that check should you assess rhythm, concision, and tone.
4. Accept changes one by one
Whole-draft acceptance hides interactions between edits. Reviewing one change at a time makes it easier to notice repeated qualifications, inconsistent terminology, or a subtle shift in emphasis. Reject or edit a proposal that solves the surface problem but changes the underlying position.
5. Read the final passage without the markup
Tracked changes are excellent for accountability but visually noisy. Before returning the document, read the accepted wording as clean prose and then inspect the tracked version. The first check protects readability; the second protects the decision record.