Background <p>Recent advancements in augmented reality (AR) have gained increasing interest in surgical applications, particularly in maxillofacial surgery. This study evaluates the accuracy and reproducibility of an AR-based navigation system using optical see-through (OST) headsets, in comparison to traditional CAD/CAM-guided approaches for Le Fort I osteotomy. Twenty identical stereolithographic skull models were divided into two groups: one group treated with OST-based AR navigation system to visualize virtual surgical lines on the skull models (<i>n</i> = 10) and the other treated with the conventional CAD/CAM procedure using titanium materialise surgical guides (<i>n</i> = 10). Virtual surgical planning (VSP) was carried out identically for both groups.</p> Results <p>Surgical accuracy was assessed by measuring deviations (in mm) between planned and postoperative positions of seven cephalometric reference points. All models in the CAD group achieved deviations below the 2-mm clinical threshold (mean deviation: 0.72 ± 0.38&#xa0;mm). However, only 4 out of 10 in the OST group met this threshold (mean deviation: 2.27 ± 1.24&#xa0;mm). A statistical analysis using Mann–Whitney <i>U</i>-test with a significance level of 0.05 was carried out to compare the mean accuracy between OST and CAD groups. The results revealed a statistically significant difference between OST and CAD groups (<i>p</i>-value &lt; 0.005). Nevertheless, progressive improvements in the OST group were observed, likely reflecting a learning curve associated with the new technology.</p> Conclusions <p>Although CAD/CAM remains more accurate, the AR-based system offers advantages in real-time visualization and reduced costs and preoperative time by eliminating the need for 3D-printed guides. However, the current accuracy limitations highlight the need for further refinement of AR systems and increased operator training. Future studies are needed to validate the clinical applicability and reliability of AR-guided orthognathic surgery.</p>