A seismic survey captures a wave field for detecting earth properties that support geological explorations. The seismic data is acquired through the use of sources and receivers located in the study field. The captured data, organized in a 3D datacube, can be incomplete because of environmental limitations, hardware malfunction, or undetonated sources to reduce costs and environmental impact. The missing data from removed sources, so-called missing shots, has been computationally recovered by compressive-based methods that usually vectorize the datacube, destroying the natural structure and redundancies in the seismic data. This work introduces a shot recovery method for a commonly used orthogonal seismic acquisition geometry, inducing a non-local self-similarity prior through the Plug-and-Play framework (PnP). The proposed method relies on the flexibility and versatility of the PnP to employ a denoiser as the promoter of the high-structural redundancies over the 3D datacube. Experiments over a synthetic-realistic dataset show the proposed method's effectiveness, obtaining improved recovery quality and reduced computational time compared to the state-of-the-art methods.