Glenohumeral osteoarthritis affects glenohumeral kinematics, but these changes in functional arm positions during upright-standing remain poorly documented. We aimed to explore the use of a biplanar radiography-based methodology for semi-static three-dimensional reconstructions of six degrees of freedom glenohumeral kinematics and to document its reliability for patients with osteoarthritis. Using computed tomography images of ten patinets with glenohumeral osteoarthritis, three-dimensional models of the humerus and scapula were reconstructed, and their anatomic coordinate systems were defined. Subjects underwent low-dose biplanar radiography in eight functional arm positions:​ relaxed standing, 45 degrees of extension, 45, 90, and 120 degrees of flexion (45F, 90F, and 120F), and abduction (45AB, 90AB, and 120AB). Two raters, with one performing a repeated measurement, used custom-software to register the three-dimensional models to biplanar image pairs for each arm position and measure the corresponding semi-static glenohumeral kinematics. Based hereon, inter- and intra-observer reliability was assessed. Our results show that translational kinematics can be measured with moderate to excellent reliability across all arm poses except for 120F and 120AB. Caution should be taken when reconstructing the joint in 90F and 90AB, as the low ICC confidence intervals indicate the need for additional attention and further observer training. Furthermore, three-dimensional glenohumeral translations were more reliable than rotations with an average ICC of 0.77 compared to 0.62, respectively. Particularly, glenohumeral internal/external rotation proved to have poor reliability. The outcomes highlight the importance of reliability assessment in patients with osteoarthritis and support future research on integrating functional measurements into total shoulder arthroplasty preoperative planning.