OBJECTIVES The aim of this study was to assess feasibility and accuracy of augmented reality-guided lumbar facet joint injections. MATERIALS AND METHODS A spine phantom completely embedded in hardened opaque agar with 3 ring markers was built. A 3-dimensional model of the phantom was uploaded to an augmented reality headset (Microsoft HoloLens). Two radiologists independently performed 20 augmented reality-guided and 20 computed tomography (CT)-guided facet joint injections each: for each augmented reality-guided injection, the hologram was manually aligned with the phantom container using the ring markers. The radiologists targeted the virtual facet joint and tried to place the needle tip in the holographic joint space. Computed tomography was performed after each needle placement to document final needle tip position. Time needed from grabbing the needle to final needle placement was measured for each simulated injection. An independent radiologist rated images of all needle placements in a randomized order blinded to modality (augmented reality vs CT) and performer as perfect, acceptable, incorrect, or unsafe. Accuracy and time to place needles were compared between augmented reality-guided and CT-guided facet joint injections. RESULTS In total, 39/40 (97.5%) of augmented reality-guided needle placements were either perfect or acceptable compared with 40/40 (100%) CT-guided needle placements (P = 0.5). One augmented reality-guided injection missed the facet joint space by 2 mm. No unsafe needle placements occurred. Time to final needle placement was substantially faster with augmented reality guidance (mean 14 ± 6 seconds vs 39 ± 15 seconds, P < 0.001 for both readers). CONCLUSIONS Augmented reality-guided facet joint injections are feasible and accurate without potentially harmful needle placement in an experimental setting.