The cracking and spalling of wet joint concrete are common forms of wet joint damage in prefabricated beam bridge structures; however, existing diagnosis methods are often insensitive to damage and are overly susceptible to environmental effects and random vehicle loads. To address this, we propose a diagnostic method for wet joint damage in such clusters on the basis of data obtained from distributed optical sensing fibers. Distributed optical sensing fibers are deployed along the top and bottom of the main girder webs, and a Brillouin optical time domain analysis (BOTDA) analyzer is used to collect strain at the corresponding locations, from which the sectional curvature is computed. By exploiting the cluster’s spatiotemporal correlation, a bidirectional long short-term memory (Bi-LSTM) network is constructed to predict the sectional curvature of bridges with identical configurations within the cluster. On this basis, a cluster-level wet joint damage diagnosis index is formulated, and a cross-validation strategy is employed to identify wet joint damage across the cluster. Under random vehicle loads, a numerical case study of wet joint damage in a prefabricated beam bridge cluster is conducted to compare the proposed approach with conventional methods, thereby verifying its effectiveness. In addition, monitoring data from three adjacent bridges in an actual cluster are used for further validation. The results indicate that the proposed method satisfies the requirements for cluster-level wet joint damage diagnosis and can be applied to the real-time monitoring and assessment of prefabricated beam bridge clusters.