Immune checkpoint therapy targeting the PD-1/PD-L1 axis is a potentially novel development in anticancer therapy and has been applied to clinical medicine. However, there are still some problems, including a relatively low response rate, innate mechanisms of resistance against immune checkpoint blockades, and the absence of reliable biomarkers to predict responsiveness. In this study of in vitro and in vivo models, we demonstrate that PD-L1–vInt4, a splicing variant of PD-L1, plays a role as a decoy in anti–PD-L1 antibody treatment. First, we showed that PD-L1–vInt4 was detectable in clinical samples and that it was possible to visualize the secreting variants with IHC. By overexpressing the PD-L1–secreted splicing variant on MC38 cells, we observed that an immune-suppressing effect was not induced by their secretion alone. We then demonstrated that PD-L1–vInt4 secretion resisted anti–PD-L1 antibody treatment, compared with WT PD-L1, which was explicable by the PD-L1–vInt4’s decoying of the anti–PD-L1 antibody. The decoying function of PD-L1 splicing variants may be one of the reasons for cancers being resistant to anti–PD-L1 therapy. Measuring serum PD-L1 levels might be helpful in deciding the therapeutic strategy.