The phenomenon of peer review reports suspected to be entirely or substantially generated by generative artificial intelligence (GenAI) is increasingly observed in scientific journal publishing. Although academic publishing institutions worldwide have issued various guidelines or statements regarding GenAI usage, these explicit regulatory documents have failed to effectively translate into substantive constraints on the behavior of reviewers. Focusing on the standardized use of GenAI in peer review, this paper analyzes the behavioral logic behind reviewers’ continued use of GenAI despite existing regulations, as well as the practical difficulties faced by editorial departments in identifying and handling such behaviors. From the perspective of daily management practices, it attempts to construct effective collaborative governance strategies. The aim is to provide references for scientific journals to resolve the practical dilemma of “regualtions without enforcement” in the GenAI era, thereby fostering a trustworthy review ecosystem.

The study reviews the public statements from various scientific journals, categorizing existing regulations into two types: “absolute prohibition” and “restriction-disclosure”. The “absolute prohibition” regulations explicitly ban GenAI involvement in the review process, including uploading manuscripts and generating review comments. The “restriction-disclosure” regulations acknowledge GenAI as an assistive tool under strict constraints, such as mandatory disclosure and prohibitions on inputting confidential content.

The findings indicate that reviewers’ use of GenAI is primarily driven by the pragmatic demands of efficiency enhancement and knowledge supplementation, compounded by a limited awareness of technological risks and insufficient deterrents against violations. Conversely, editorial departments face a quadruple dilemma when enforcing GenAI regulations: First, a lack of technical means to monitor reviewers’ local operations on manuscripts, leaving confidentiality unprotected; Secondly, the difficulty of forensically verifying AI-generated content, complicating the quality assessment of reviews; Thirdly, the absence of objective criteria for identifying violations, making qualitative judgements sensitive and contentious. Fourthly, limited human and financial resources to conduct specialized verification.

The existing “absolute prohibition” or “restriction-disclosure” regulations fail to effectively reconcile the practical needs of reviewers with the compliance bottom line of journals, significantly diminishing their practical efficacy. The study posits that addressing the use of GenAI in peer review requires moving beyond mere prohibitions toward establishing a new type of collaborative relationship based on shared goals and mutual understanding. Editorial departments should not only act as regulators but also supporters and enablers of responsible reviewing. Accordingly, the paper proposes a four-point framework for standardized use: formulating clear and operable “Guidelines for Reviewers on the Use of GenAI Auxiliary Tools”; embedding normative requirements into the online submission and review workflow; exploring and providing institutionally deployed, secure auxiliary tools; and establishing a reviewer credit record and positive incentive mechanism. This approach aligns the individual rational choices of reviewers (pursuing review efficiency and accumulating academic reputation) with the collective goals of the journals (ensuring journal quality and maintaining academic integrity). Through a governance model combining “guidance + support + incentive”, scientific journals can achieve a dual improvement in both the quality and efficiency of peer review while safegarding academic integrity.