The vertebral fracture patterns of AO classification have been established historically via radiograph and computed tomography analysis, achieving modest reproducibility values. The authors hypothesize that magnetic resonance imaging may improve reliability because it better indicates posterior ligamentous complex damage. They conducted a retrospective analysis of a prospective recruited cohort of patients using radiographs and magnetic resonance images with fat saturation sequences to classify 37 traumatic vertebral fractures. Five spine surgeons, 2 orthopedic residents, 2 musculoskeletal radiologists, and 2 radiodiagnosis residents classified the morphological pattern of each fracture per AO classification in 2 separate sessions that occurred 6 weeks apart. Inter- and intraobserver reproducibility for AO classification types A, B, and C were assessed using the kappa test (pairwise method), and standard error was assessed using the jackknife method. Quantitative comparisons were performed using the Student's t test, and the kappas were performed using normal approximation. Mean interobserver agreement was kappa=0.53 and 0.47 for the first and second sessions, respectively, for all evaluators. Greater interobserver agreement was observed between the senior doctors (kappa=0.59 and 0.54 for the first and second sessions, respectively) vs residents (kappa=0.45 and 0.31 for the first and second sessions, respectively) (P=.02) and between orthopedic surgeons vs radiologists (kappa=0.71 vs 0.48, respectively) (P=.008). Mean intraobserver agreement was kappa=0.58 (range, 0.38-0.76). Evaluators more familiar with the classification obtained higher kappas. Magnetic resonance imaging offers moderate reproducibility in assessing vertebral fractures pursuant to AO classification, and results are slightly better than those reported with computed tomography. Reliability increases in the hands of experiened spine surgeons and improves with greater familiarization with the classification.