Abstract Purpose: Immune checkpoint inhibitors (ICI) are effective in only a minority of patients with esophagogastric cancer (EGC). Here, we aimed to identify predictors of durable clinical benefit to ICI in EGC. Experimental Design: Patients with advanced EGC treated with ICIs at Memorial Sloan Kettering Cancer Center (New York, NY) were identified. Clinicopathologic variables were assessed. In patients profiled by MSK-IMPACT–targeted sequencing, outcomes were correlated with tumor genomic features. Results: One-hundred sixty-one patients were treated with ICIs (110 with anti–PD-1/PD-L1 antibodies and 51 with anti–CTLA-4 and PD-1/PD-L1 antibodies). The median progression-free survival (PFS) and overall survival (OS) were 1.7 and 4.9 months, respectively. Greater number of disease sites (≥3), liver metastases, treatment with ≥3 prior therapies and ECOG performance status ≥2 were associated with poorer PFS and OS. Patients treated with combination ICI and those with PD-L1–positive tumors had improved outcomes. There was no difference in outcomes between patients treated with antibiotics during or in the 2 months preceding ICI treatment versus those who were not. Occurrence of irAEs was associated with improved OS. In genomically profiled tumors (n = 89), survival was associated with increasing tumor mutation burden (TMB). However, in multivariable analyses and when microsatellite unstable (MSI) patients were excluded, a significant association was no longer observed. Conclusions: In patients with advanced EGC, heavily pretreated patients, those with high-volume disease and/or poor PS were less likely to benefit from ICI. irAEs were associated with improved OS. TMB correlated with improved survival, but this association was not observed when MSI-high patients were excluded.