BACKGROUND: High-quality cardiopulmonary resuscitation (CPR) is critical to cardiac arrest patients. Extended Reality (XR) technologies, including Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR), provide immersive and interactive training, potentially enhancing CPR outcomes. This network meta-analysis compared the effectiveness of XR-based CPR training to traditional face-to-face methods. METHODS: A Bayesian network meta-analysis was conducted following PRISMA guidelines. We systematically searched PubMed, Cochrane Library, Web of Science, EMBASE, and CNKI for randomized controlled trials (RCTs) comparing XR-based and traditional CPR training. Primary outcomes included chest compression depth and rate; secondary outcomes assessed full chest wall recoil. The CINeMA tool (GRADE framework) was used to assess evidence quality. Statistical analyses were performed using Stata 15 SE and ADDIS software with random-effects models. RESULTS: 11 RCTs (1,190 participants) were included. MR showed the improvement in chest compression depth (SMD = 10.96; 95% CI, 0.95 to 20.82) compared to VR and traditional methods. For full chest wall recoil, AR outperformed VR (SMD = 48.57; 95% CI, 19.56 to 79.75) and traditional methods (SMD = 52.95; 95% CI, 25.94 to 80.48). However, no significant differences were observed for chest compression rate. SUCRA rankings placed MR as most effective for compression depth (87.4%) and AR for full chest wall recoil (99.1%). Evidence quality was moderate to high, with minor downgrades for imprecision. No publication bias was detected. CONCLUSIONS: XR technologies, particularly MR and AR, significantly improve chest compression depth and full chest wall recoil in comparing with face to face CRP training, offering a flexible and engaging approach to CPR training. Further studies are needed to evaluate long-term skill retention and real-world impact. CLINICAL TRIAL NUMBER: Not applicable.