How to apply various kinds of artificial intelligence technology in the publishing field and how to reconstruct the publishing implementation process with large models have become the focus of the publishing industry and academia. Book quality is not only the bottom line of book publishing, but also the lifeline of publishing house. So how to improve the efficiency and accuracy of the review work through the intelligent review system, ensure the quality of book publishing, and help the publishing unit to hold the line of defense has become an industry topic that needs to be focused on. From the three stages of review, editing, processing and collation, the paper analyzes the pain points of manual review, the ways of intelligent review to solve the pain points of manual review, and the problems difficult to solve by intelligent review, and further proposes the man-machine collaborative proofreading mode of “manual review + intelligent review”: In the review stage, “preliminary intelligent review + manual review”, in the editing processing and finishing stage, “intelligent assisted editing + manual fine processing”, in the school review stage, “intelligent proofreading + manual final review” and feedback optimization, so as to help the publishing unit to hold the line of defense and ensure the quality of book publishing.