Objectives: This study aimed to determine interobserver reliability and intraobserver reproducibility of the OTA/AO classification for humeral shaft fractures, and to evaluate differences between fracture types, fracture groups, and surgical specializations. Methods: Thirty observers (25 orthopaedic trauma surgeons and 5 general orthopaedic surgeons) independently classified 90 humeral shaft fractures according to the OTA/AO classification. Patients of 16 years and older were included. Periprosthetic, recurrent, and pathological fractures were excluded. Radiographs were provided in random order, and observers were blinded to clinical information. To determine intraobserver agreement, radiographs were reviewed again after 2 months in a different random order. Agreement was assessed using kappa statistics. Results: Interobserver agreement for the 3 fracture types was moderate (κ = 0.60; 0.59-0.61). It was substantial for type A (κ = 0.77; 0.70-0.84) and moderate for type B (κ = 0.52; 0.46-0.58) and type C fractures (κ = 0.46; 0.42-0.50). Interobserver agreement for the 9 fracture groups was moderate (κ = 0.48; 95% CI, 0.48-0.48). Orthopaedic trauma surgeons had better overall agreement for fracture types, and general orthopaedic surgeons had better overall agreement for fracture groups. Observers classified 64% of fractures identically in both rounds. Intraobserver agreement was substantial for the 3 types (κ = 0.80; 0.77-0.81) and 9 groups (κ = 0.80; 0.77-0.82). Intraobserver agreement showed no differences between surgical disciplines. Conclusions: The OTA/AO classification for humeral shaft fractures has a moderate interobserver and substantial intraobserver agreement for fracture types and groups.