Glioblastoma (GBM) is an aggressive brain tumor marked by extensive heterogeneity, resistance to therapy, and dismal prognosis. Extracellular vesicles (EVs) have emerged as key players in GBM biology, mediating intercellular communication and therapy adaptation. However, the exact functions, significance, and molecular impact of EVs in GBM are still unclear. In this study, we performed a comparative proteomic analysis of U87 GBM cells grown in two-dimensional (2D) monolayers and three-dimensional (3D) spheroids following temozolomide (TMZ) treatment, alongside the characterization of EVs derived from both types of cell cultures. 3D spheroids secreted more EVs of smaller size and exhibited a more TMZ-resistant, stem-like proteome under the TMZ-induced genotoxic stress. In contrast, 2D cells showed greater proteomic remodeling, with EVs enriched in protein families involved in DNA repair, oxidative stress adaptation, and methylation processes. Notably, several methyltransferases were downregulated intracellularly but selectively retained in EVs, suggesting active sorting to influence the tumor microenvironment or modulate epigenetic states in recipient cells. EVs also carried adhesion molecules and signaling proteins linked to migration, invasion, and Wnt pathway activation, as well as metabolic enzymes connecting serine metabolism and redox control to TMZ resistance. Mapping of the EV and cellular proteomes onto The Cancer Genome Atlas (TCGA) dataset identified prognostic protein families associated with either poor or favorable patient outcomes. Together, our data demonstrate how EV cargo composition reflects TMZ adaptation and tumor phenotype, offering insights into mechanisms of resistance to TMZ and potential biomarkers for patient stratification and therapeutic targeting in GBM.