Review Response To Reviewers

/review-response-to-reviewers

This letter is read by the same reviewers who wrote the comments, and they check one thing above all: was every point addressed? A missed point guarantees another review round, and the process can already run two or three times.

The highest-value check here is not stylistic. It is verifying that changes claimed in the letter actually exist in the manuscript.

Before anything else

1. Load the style rules. Invoke paper-writing-style. If absent, use the condensed rules in create-response-to-reviewers.
2. Lock the language. Language of the user's first message. Rewrites are English.
3. Get all three documents if possible: the reviewer comments, the response letter, and the revised manuscript. Each check below needs a different pair, and the most important one needs all three.

Check 1: coverage

Re-enumerate the comments from the original reviews, independently of how the letter numbered them. Reviewer prose hides multiple requests in single paragraphs, and a letter that numbered them generously may still have merged two.

Build a coverage table: comment number, the comment's substance in a few words, whether the letter responds, and whether a manuscript location is given.

Report:

1. Unanswered comments. The most serious finding. Quote the comment and say it has no response.
2. Partially answered comments, where a two-part request got a one-part answer. Very common with "Furthermore, the authors should also" sentences.
3. Responses with no manuscript location. A reviewer will not go looking. Every accepted change needs a section, paragraph, page or line.
4. Responses that answer a different question than the one asked.

Check 2: claimed changes actually exist

For every change the letter claims, look for it in the revised manuscript.

Report any claim you cannot verify, naming the claim and the location it cites. Distinguish clearly between "the change is not at the cited location" and "I could not find it anywhere", since the first is a numbering error and the second is far more serious.

A response claiming a change that was not made is the worst outcome available here. The reviewers have the manuscript. Flag every unverifiable claim prominently, and do not soften it.

If the revised manuscript is not available, say plainly that this check could not be run and that it is the one worth running before sending.

Check 3: classification honesty

For each response, judge whether the class matches the content:

• A response classed as accepted that actually declines the point, hedged into vagueness. Reviewers read this as evasion and it is worse than an honest refusal.
• A point declined without a reason.
• A "clarification" response that does not clarify, or that blames the reviewer for misreading without noting that the text was made clearer.
• Twenty comments all accepted with no pushback anywhere. Sometimes correct, sometimes a sign of a paper being weakened to please reviewers. Worth a single line, not a lecture.

Check 4: tone

The letter is read by people who will decide the paper's fate. Flag:

• Any implication that the reviewer was careless, inattentive, or wrong to ask.
• Defensiveness, or arguing at length where a short explanation and a text clarification would do.
• Effusive thanks on every comment. Once, where a comment genuinely improved the paper, reads as sincere. Twelve times reads as filler.
• Sarcasm or irritation, in any quantity.

Quote the sentence and give a neutral rewrite. Tone problems in a response letter cost papers, and the author is usually the last person able to see them.

Check 5: structure and style

• Are comments quoted verbatim? The reviewer needs to find their own point without reopening their review.
• Is the reviewer attributed for each comment, and numbering consistent?
• Is the accompanying editor cover letter present, stating that the comments were addressed and that changes are highlighted in the revised manuscript?
• Does the letter mention that changes are highlighted? Reviewers rely on this.
• Style: em dashes, vague language, nominalization, AI tells.

Deliver

1. The coverage table first. It is the finding that matters.
2. Unverifiable claimed changes, prominently.
3. Classification and tone findings, each with the sentence quoted and a rewrite.
4. Structure and style hits.
5. Rewrites of the specific responses that failed, not the whole letter. Response letters are long and most responses are usually fine.

If everything checks out, say so and say explicitly that you verified the claimed changes against the manuscript, since that is the reassurance worth having before sending.

Hard rules

• Never soften a finding that a claimed change is missing.
• Never invent a manuscript location to fill one that is absent.
• Never invent a reason for a declined point.
• Never write a response to a comment the researcher has not addressed. Report the gap; the answer is theirs.
• Never emit an em dash.
• Say which checks you could not run if a document is missing.