Artificial intelligence in shoulder arthroplasty: latest concepts and clinical integration

BACKGROUND: Recent advances in artificial intelligence (AI) have redefined shoulder arthroplasty, ultimately improving diagnostic accuracy, refining surgical planning, and personalizing recovery pathways. However, despite promising technical applications, comparative clinical evidence has not yet demonstrated improved routine surgical or patient outcomes in shoulder arthroplasty. REVIEW CONTENT: Therefore, this review summarizes the state-of-the-art AI applications in shoulder arthroplasty, highlights their strengths and limitations, and outlines integration strategies that align technological innovation with surgical expertise. Preoperatively, deep learning-based imaging, which refers to machine learning methods that use multilayer neural networks to identify complex patterns in imaging data, has shown strong diagnostic performance in detecting rotator cuff tears and glenohumeral arthritis, while AI-assisted three-dimensional reconstruction and anatomy-based modeling help visualize glenoid deformity and support more intuitive preoperative planning, complementing conventional navigation. Intraoperatively, AI-based phase recognition and robotic assistance may enhance component placement precision while also offering potential benefits in surgical training, workflow standardization, and real-time feedback. In addition, emerging innovations such as mixed reality guidance, multimodal integration, and semi-autonomous assistance continue to expand the potential scope of AI applications in shoulder arthroplasty. Postoperatively, machine learning-driven prognostic models and wearable monitoring systems predict complications and guide tailored rehabilitation, supporting earlier risk identification when required. However, current AI systems are unable to capture surgeon intuition and context-dependent expertise. Nevertheless, emerging strategies such as explainable AI, multimodal integration, federated learning, and semi-autonomous surgical platforms aim to bridge these gaps by enhancing transparency, scalability, and clinical trust. CONCLUSION: The adoption of AI tools by surgeons depends on their demonstrated reliability across diverse patient anatomies, intraoperative variability, and real-world rehabilitation settings. Thus, technical refinement along with mutual trust with surgeons, supportive regulatory frameworks, and institution-specific implementation strategies are essential for safe integration, particularly because current systems may increase cost, workflow complexity, or delay care when not adequately validated or appropriately implemented.