In several malignancies, only a limited number of patients respond to immune checkpoint inhibitors. Predicting and monitoring responses to these inhibitors represent an unmet clinical need. Here, we developed a PET/CT probe targeting granzyme B, [68Ga]Ga-NOTA-Gly-Gly-Gly-Ile-Glu-Pro-Asp-CHO (GSI), and aimed to investigate whether it can be used to monitor the effects of immune checkpoint inhibitors early in the course of therapy. Methods: Seventy-two patients with gastric cancer (stages III-IV) were recruited for [68Ga]Ga-DOTA-GSI PET/CT imaging after 2 or 3 cycles of the immunotherapy, and 40 patients were included in the final analysis. The SUVmax of primary tumors (SUVmax-t), SUVmax of metastatic lymph nodes (SUVmax-LN), and SUVmax of normal tissues (liver and blood pool) were measured, and their target-to-liver background ratio (TLR) and target-to-blood background ratio (TBR) were denoted for primary tumors as TLRtumor and TBRtumor and for metastatic lymph nodes as TLRLN and TBRLN, respectively. The treatment responses were assessed within 1 wk after full-course treatment according to RECIST version 1.1. Wilcoxon rank-sum tests were used to compare the PET/CT parameters between responders and nonresponders. Receiver operating characteristic curve analysis was used to assess the diagnostic efficacy of [68Ga]Ga-DOTA-GSI PET/CT parameters in identifying responders. Two-tailed P value of less than 0.05 was considered statistically significant. Results: We found that SUVmax-t, TLRtumor, TBRtumor, SUVmax-LN, and TBRLN were higher in responders than in nonresponders (2.49 ± 0.58 vs. 1.55 ± 0.48, P = 0.000; 2.24 ± 0.48 vs. 1.74 ± 0.67, P = 0.007; 1.38 ± 0.43 vs. 0.90 ± 0.23, P = 0.000; 2.24 ± 0.99 vs. 1.42 ± 0.55, P = 0.003; and 1.28 ± 0.68 vs. 0.83 ± 0.32, P = 0.012, respectively). According to receiver operating characteristic curve analysis, the area under the curve for SUVmax-t, TBRtumor, TLRtumor, SUVmax-LN, TLRLN, and TBRLN was 0.886, 0.866, 0.746, 0.772, 0.648, and 0.731, respectively. The threshold of SUVmax-t was 2.05, and its sensitivity and specificity were 81.0% and 84.2%, respectively. In addition, multivariate logistic regression indicated that TBRtumor was an independent predictor of treatment response (P = 0.03). Conclusion: Our results indicated that [68Ga]Ga-DOTA-GSI PET/CT is a promising tool for predicting early response to combined immunotherapy in gastric cancer patients.