Protein glycosylation plays crucial roles in the regulation of diverse biological processes. As a critical step, mass spectrometry-based site-specific analysis of protein glycosylation is important to better understand these events. Despite the great progress, characterization of structural isomers of glycans and glycopeptides remains challenging. In typical glycoproteomic analysis, collision-induced dissociation (CID) or higher-energy collisional dissociation (HCD) fragmentation produces abundant saccharide oxonium ions containing N-acetylhexosamine (HexNAc) residues. However, it has been difficult to distinguish isobaric GalNAc and GlcNAc modifications by using mass spectrometry only. By using intensities of oxonium ions of standard O-GlcNAc/O-GalNAc peptides, we systematically investigated the fragmentation patterns of different ions. Then a binary logistic regression model was established by training comprehensive data sets from glycoproteomics studies reported. The model was then tested with independent O-glycoproteomics data sets, with reliable classification achieved (>87% accuracy). In comparison to empirical observations and criteria used previously, our model is accurate and generalized. Based on this model, a corresponding Web server HexNAcQuest has been constructed, which is freely accessible to users. The model can also be easily integrated in MS-based glycoproteomics workflows to distinguish the isobaric HexNAc modifications.