Abstract Objective After temozolomide failure, no evidence-based treatment is available for pituitary carcinomas (PCs) and aggressive pituitary tumors (APTs). To date, only 12 cases treated with immune-checkpoint inhibitors (ICIs) have been published, showing encouraging efficacy. Predictive factors of response are lacking. Here, we aimed to assess the real-life efficacy and predictors of response to ICIs in PCs and APTs. Design and methods This study is a multicentric, retrospective, observational cohort study, including all PCs and APTs treated with ICIs in France up to March 2022. PD-L1 immunohistochemistry and CD8+ T cell infiltration were evaluated centrally. Results Six PCs (four corticotroph and two lactotroph) and nine APTs (five corticotroph and four lactotroph) were included. The real-life efficacy of ICIs was lower than previously published data. Three corticotroph tumors (33.3%) showed partial response, one (11.1%) stable disease, while five (55.6%) progressed. One lactotroph tumor (16.7%) showed partial response, one (16.7%) stable disease, while four (66.7%) progressed. PCs responded far better than APTs, with 4/6 PCs showing partial response compared to 0/9 APTs. Corticotroph tumors responded slightly better than lactotroph tumors. In the four responsive corticotroph tumors, PD-L1 staining was negative and CD8+ T cell infiltration attained a maximum of 1% in the tumor center. Conclusions Confirmation of the presence or absence of metastases is necessary before starting ICIs. After temozolomide failure, ICIs appear as a good therapeutic option for PCs, especially for corticotroph carcinomas. Negative PD-L1 staining and very low CD8+ T cell infiltration in the tumor center should not preclude ICI administration in corticotroph carcinomas. Significance statement This is the first study to assess the real-life efficacy of ICIs in pituitary carcinomas (PCs) and aggressive pituitary tumors. We also assessed potential predictors of response and are the first to assess the predictive value of CD8+ cell infiltration. We identified the tumor type as a major predictor, ICIs proving far more effective in treating PCs. Our study provides evidence that ICIs are a good option after temozolomide failure for PCs (four of six responded), especially for corticotroph carcinomas (three of four responded). We also provide evidence that negative PD-L1 staining and very low CD8+ cell infiltration in the tumor center should not preclude ICI administration in corticotroph carcinomas. Moreover, our findings point toward the need to systematically perform extension workup before starting ICIs.