BACKGROUND: Glenohumeral (GH) joint steroid injection is one of the most well-known treatments for frozen shoulder. However, the low accuracy of GH joint injections and the improvement of symptoms after subacromial (SA) steroid injections led us to design a study that compares the efficacy of a steroid injection for primary frozen shoulder according to the injection site. MATERIALS AND METHODS: Patients with primary frozen shoulder were randomly divided into 2 groups according to the location of the injection: a GH group of 37 for the glenohumeral joint and an SA group of 34 for the subacromial space. Injections were completed using ultrasonographic guidance. Evaluations using a visual analog scale (VAS) for pain, the Constant score, and passive range of motion (ROM) were completed at 3, 6, and 12 weeks after the injection. RESULTS: The GH group showed lower pain VAS at 3 weeks, but no statistical difference was found between the 2 groups at 6 and 12 weeks. Improvement in pain was evident at every follow-up visit compared with the preinjection evaluation. There was no significant difference between the 2 groups with respect to the Constant score or ROM at serial follow-up. CONCLUSIONS: The GH steroid injection was not superior to a SA injection for patients with primary frozen shoulder even though injection at the GH joint led to earlier pain relief compared with the SA injection. SA steroid injection along with a GH injection is an alternative modality, and the treatment should be individualized and tailored appropriately.