Bacteraemia, the presence of pathogens in the bloodstream, is associated with long-term morbidity and is a potential precursor condition to life-threatening sepsis. Timely detection of bacteraemia is therefore critical to reduce patient mortality, but existing methods lack the precision, speed, and sensitivity to effectively stratify bacteremic patients from uninfected individuals and identify the infecting pathogen. Herein, we test the potential of quantitative serum N-glycomics performed using an established porous graphitised carbon LC-MS/MS platform to stratify bacteremic patients infected with different pathogens i.e. Escherichia coli, Staphylococcus aureus (both n = 11), Pseudomonas aeruginosa, and Streptococcus viridans (both n = 5) from healthy donors (n = 39). In total, 62 N-glycan isomers spanning 41 glycan compositions primarily comprising complex-type core fucosylated and α2,3- and α2,6-sialylated structures were profiled across all samples using label-free relative quantitation. Excitingly, unsupervised hierarchical clustering of the serum N-glycome data accurately separated the patient groups. P. aeruginosa-infected patients displayed the most aberrant serum N-glycome involving elevated core fucosylation and reduced α2,6-sialylation relative to healthy donor sera as further supported by quantitative lectin blotting using Aleuria aurantia lectin (AAL) and Sambucus nigra lectin (SNA), respectively. Importantly, receiver operating characteristic analyses demonstrated that a single N-glycan isomer can effectively stratify each of the four patient groups from the healthy donor cohort (ROC 0.93-1.00). We conclude that the serum N-glycome represents a new hitherto unexplored class of potential diagnostic markers for bloodstream infections.