Glycosylation, a common post-translational modification of proteins, plays a role in numerous biological processes. However, the role of glycosylation in colorectal cancer (CRC) remains incompletely understood. In this study, we identified that CAFs exhibited the highest glycosylation levels in CRC. We classified 6 CAFs subgroups with distinct glycosylation profiles, revealing notable heterogeneity in functional activities, communication pathways, developmental trajectories, and metabolic states. Furthermore, we developed and validated a robust prognostic model capable of predicting CRC patient survival outcomes using 101 machine learning algorithms. The model stratified patients into high-risk and low-risk groups, where genetic and epigenetic alterations, immune infiltration patterns and responses to various therapeutic drugs varied significantly between the groups. In vitro experiments demonstrated that the key gene CCDC85B in the model influences the glycosylation levels of CRC and CD8 + T cell infiltration, underscoring its potential as an essential therapeutic target. These findings underscore the functional complexity within CAFs subgroups and highlight potential therapeutic targets for CRC treatment. These findings deepen our understanding of glycosylation and offer tools for prognosis, drug selection, and targeted therapy in CRC patients.