Emerging neurostimulation methods aim to selectively modulate deep brain structures. Guiding these therapies has presented a substantial challenge, since imaging modalities such as MRI limit the spectrum of beneficiaries. In this study, we assess the guidance accuracy of a neuronavigation method that does not require taking MRI scans. The method is based on clearly identifiable anatomical landmarks of each subject's face. We compared this technique to the ideal case, MRI-based nonlinear brain registration, and evaluated the accuracy of both methods across ten targets located in deep brain structures: 7 targets in the anterior cingulate cortex as well as the anterior commissure and posterior commissure. Compared with the ideal case, the average localization error of the MRI-free method was 5.75 ± 2.98 mm (mean ± sd). These findings suggest that this method may provide sufficient compromise between practicality and the accuracy of targeting deep brain structures.