Transcranial magnetic stimulation (TMS) is a noninvasive brain stimulation technique that uses magnetic pulses to safely modulate neural activity in specific brain regions, with particular use in the treatment of major depressive disorder. However, effective treatment protocols require multiple sessions over time and accurate targeting. Current neuronavigation systems can improve pulse delivery compared to standard scalp and measuring tape methods, but are costly and have time-intensive setups for recurring sessions. We present an augmented reality neuronavigation system (AR-NS) that overcomes these limitations. The AR-NS is functionally similar to other neuronavigation systems but can operate entirely within a head-mounted display. Unlike traditional 2D navigation systems that require mentally fusing screen information with real-world actions, the AR interface allows operators to perform the procedure with reduced cognitive and hand-eye coordination demands. We measured the functional targeting accuracy of the AR-NS and the Localite TMS Navigator, a commercial neuronavigation system, using a TMS phantom we developed embedded with Hall effect sensors at four different sites. We determined the coil placement for each site that maximized the sensor response to a magnetic pulse using co-registered X-ray computed tomography and structured light scanner scans. A MagVenture C-B60 coil was placed and fired at each site 30 times in a randomized order according to each neuronavigation system. We measured the resulting magnetic pulse amplitudes. A non-inferiority test with a 5 mT margin and 97.5% confidence intervals indicated that the AR-NS demonstrated similar functional targeting accuracy as the Localite TMS Navigator across all but one stimulation site, indicating that augmented reality neuronavigation systems may offer more accessible delivery of TMS stimulation without sacrificing functional accuracy compared to current systems.