Peer review has long served as a core mechanism for quality control and resource allocation in science. However, amid the rapid growth of research output, the traditional peer review system is increasingly strained in terms of efficiency, cost, and fairness, raising concerns about its institutional capacity. The rapid development of artificial intelligence (AI) has led to its gradual incorporation into academic review processes, where it is increasingly regarded as a potential institutional variable capable of reshaping conventional peer review. Building on a historical analysis of the evolution of peer review, this article examines the institutional legitimacy of AI's involvement in peer review and the practical drivers behind its adoption. While AI demonstrates potential advantages in accelerating review workflows, optimizing the allocation of review resources, and assisting in the correction of evaluation outcomes, its integration may also generate systemic risks, including the weakening of scholarly judgment, the amplification of algorithmic bias, the blurring of responsibility boundaries, and the encroachment of efficiency−oriented logic upon academic ethics. This article argues that the key to integrating AI into peer review lies in the careful delineation of its functional boundaries and governance frameworks. It further suggests that conceptualizing AI as a constrained auxiliary tool, and applying it cautiously within a human–machine collaborative model, may help enhance review efficiency and procedural fairness while mitigating the institutional risks associated with AI involvement in academic evaluation.