You've spent two weeks on the revision. The manuscript is tighter, the limitations section expanded, the figure buried in the appendix moved to the main text where Reviewer 1 asked for it. The response letter is four pages and addresses each comment in sequence.
The question is whether you actually addressed the concerns, or just wrote four pages next to them.
Response Review reads three things together: the original reviewer comments, your draft response letter, and the revised manuscript. The reason it needs all three is that a response letter can't be checked in isolation. A comment about a missing control gets addressed in the manuscript. A concern about statistical framing gets addressed in the letter. If you write "we added a robustness check," the tool can verify whether that check is present in the revision.
The input is plain text for the reviewer comments and response letter, plus a PDF for the manuscript. Common comment formats are auto-detected. The review takes three to five minutes for a typical set.
What you get back is organized per comment, not per tool pass. For each reviewer concern, three voices weigh in: a Skeptical Reviewer who asks whether the response actually engages the critique or sidesteps it, a Tone Editor who flags defensive or dismissive phrasing, and an Editor-in-Chief who estimates whether this individual response would pass.
That per-comment structure matters. When you have thirty reviewer comments spread across three reviewers, a summary assessment doesn't tell you which one will come back again. You need to see which specific responses got flagged and why.
A few things the tool reliably catches: responses that promise a fix without citing where in the manuscript it appears; responses that restate your original argument rather than engage the critique; and tone that reads as defensive to a third party even when you read it as confident. The last one is easy to develop without noticing. You've been looking at these comments for two weeks and your frustration with Reviewer 2's apparent misunderstanding has probably made it into the letter somewhere.
There's also an overall accept/revise estimate at the end, which is useful mainly as a sanity check. If the estimate is "likely major revisions again," it's a signal to read through the flagged comments carefully before you send.
The cost is $6. Once you retrieve the results, the reviewer comments, letter, and manuscript are deleted from the servers.
The actual value is the gap between what you meant to say and what a reviewer will hear. That gap is hard to measure when you've written the same letter four times.