Few moments in a researcher’s life are as charged as opening the editor’s decision email. “Major revision” is not a rejection. It is an invitation. But whether that invitation turns into an acceptance depends less on the revised manuscript than most authors think, and far more on one document they often write last and fastest: the response to reviewers.

It helps to remember who is on the other side. Almost no reviewer is paid. They review out of scientific interest, and many do it in their own free time, after their day’s work is done. When such a reviewer opens a response letter that is clear, complete, and easy to navigate, they re-read your paper in a generous frame of mind. When they open one that is disorganised, where they have to jump back and forth to work out which point you answered and how, it puts them in a bad mood. And it does something more dangerous than that.

A messy response signals that the author is disorganised, that they cannot even write a simple, structured letter. From there it is a short step to doubting the work itself. Even when the science is sound, a reviewer who decides the author is careless may stop trusting the manuscript. That is the real risk. Not one rejected paper, but a reputation for not working rigorously, which can follow you and quietly damage a career. The response letter, in other words, is part of how your competence is judged.

Here is the approach I use and recommend.

1. Start with the right mindset

Three principles change everything before you write a single line.

2. Use a strict point-by-point structure

Editors and reviewers should never have to hunt for your reply. Every choice you make in the letter should serve one goal: to let the reviewer see, with the least possible effort, that each of their comments has been understood and acted on.

Use a point-by-point letter, and give every comment the same three-part shape, in the same order (Figure 1).

1 · Reviewer comment — quoted verbatim (shown in grey)
2 · Your response — in black: what you did, and the concrete result
3 · The exact change in the manuscript — quoted verbatim, with its location
And always: deliver the revised manuscript with tracked changes switched ON (non-negotiable)
Figure 1 · The anatomy of one point-by-point entry.

Three things make this work. First, quote the reviewer’s comment verbatim. It proves you read it precisely, and it lets a busy reviewer check your reply without reopening the original review. Part of quoting well is deciding what counts as one point: a reviewer often packs several independent issues into a single sentence, and answering them in one lumped reply is exactly how a point quietly gets lost. Break the review into logical units, one entry per issue, so that nothing slips through. Second, give your response: what you changed, and the concrete result. Third, repeat verbatim what actually changed in the manuscript: the exact sentence you removed and the exact sentence you added, with the page and line. Some of this also appears in your response, but repeating it here gives the reviewer maximum transparency and saves them from switching between the letter and the manuscript.

Two small habits make the letter dramatically easier to read. Set the reviewer’s quoted comment in grey and keep your own response in black, so the two are separated at a glance. And deliver the revised manuscript with tracked changes turned on. This is non-negotiable. It lets the reviewer confirm every edit in seconds rather than hunting for it. These are small things, but they show that you work in a structured way, that you have done everything you can to make the reviewer’s job easy, and that you have nothing to hide.

Example: a strong reply to a tough comment

Reviewer comment: “The authors claim biomarker X predicts progression, but the analysis does not adjust for tumour grade, which is a known confounder.”

Response: We thank the reviewer for raising this important point. We agree that tumour grade is a potential confounder. We have repeated the analysis with a multivariable Cox model adjusting for grade, age, and treatment arm. Biomarker X remained independently associated with progression (adjusted HR 1.8, 95% CI 1.2 to 2.7, p = 0.004). We have added this analysis to the Results and a new Table 3, and we have tempered the causal language in the Discussion.

Change in the manuscript (verbatim): In the Discussion (p. 12) we replaced “biomarker X predicts progression” with “biomarker X is independently associated with progression after adjustment for tumour grade.”

Notice the shape of the reply. You acknowledge the point, agree where you honestly can, do the work, report the concrete result, point to the exact location, and soften the claim. A reviewer can close a comment like that in under a minute.

3. How to disagree with a reviewer (without losing)

You will not agree with everything, and you should not pretend to. But disagreement has to be evidence-led and courteous, never dismissive. One rule sits above all the others: never make the reviewer look foolish.

Even when a reviewer is plainly wrong, correct the point once, plainly, and without any edge. No sarcasm, no “as the reviewer surely knows,” nothing that an irritated reader could take as an insult between the lines. A reviewer who feels belittled becomes an adversary, and an adversary can sink a sound paper out of sheer ill will. Reviewers are people too, with egos and off days, and a courteous reply gives them every reason to stay on your side. Good manners here are not optional politeness. They are strategy.

Example: declining an experiment, respectfully and thoroughly

Response: We appreciate this suggestion. A knockout validation would indeed strengthen mechanistic inference. Generating and characterising the model, however, is a multi-month effort that falls outside the scope of this clinical-correlative study. To address the underlying concern as fully as we can, we have (a) added orthogonal evidence from the published literature (refs 18 to 20), (b) re-analysed our existing data to test the same hypothesis a second way, and (c) explicitly noted the absence of functional validation as a limitation (Discussion, p. 13). We would be glad to pursue the knockout work in a dedicated follow-up study.

The reviewer and the editor should come away certain that you did everything reasonable to resolve the point. That impression of genuine effort is often worth as much as the experiment itself.

4. Handle the hard cases

Contradictory reviewers

When Reviewer 1 wants more detail and Reviewer 2 wants brevity, do not try to satisfy both in the same sentence. Make a judgement, implement it, explain your reasoning to both, and flag the conflict to the editor, whose call it ultimately is.

“More experiments”

Separate the requests you can do quickly from those you cannot. Do the quick, high-value ones. For the rest, give the scope-and-limitation answer above. A revision is not an obligation to do unlimited work. It is an obligation to make the paper sound. And if you do produce new data or figures, hold them to the same standard the journal expects of everything else. A figure made only because a reviewer asked for it is still your figure, and a sloppy one quietly undoes the goodwill you were working to earn.

A comment you genuinely do not understand

Do not guess. Re-read the passage as the reviewer would. If it is still unclear, answer the most reasonable interpretation and say so openly. Often the very fact that a comment was unclear means your text was unclear too, so revise it.

5. Get the tone right

Gratitude, not grovelling. Confidence, not combativeness. A few habits consistently land well.

6. A pre-submission checklist

Done well, a response letter does something subtle. It turns the reviewer from a gatekeeper into a collaborator on the final version, someone who wanted to help all along. That shift is what turns “major revision” into “accept.”