<p>Fracture healing is a dynamic multiscale process jointly driven by mechanical stimuli, cellular behaviors, and biochemical regulation. Computational models have played an important role in elucidating healing mechanisms, evaluating fixation strategies, and optimizing rehabilitation interventions. However, their increasing structural complexity has also introduced challenges, including limited parameter identifiability, high computational cost, and insufficient patient-specific adaptability. In recent years, artificial intelligence (AI) methods have gradually been introduced into fracture-healing modeling studies, providing new methodological pathways for extending traditional mechanism-based models. In the form of a narrative review, this article examines the convergence of computational models and AI methods in fracture healing. It summarizes recent advances in mechanoregulatory models, bioregulatory models, and mechanobioregulatory models, with particular emphasis on the application of AI in fracture-healing simulation, patient-specific modeling, and extensions to bone tissue engineering. Furthermore, this review discusses the key bottlenecks encountered in translating fracture-healing computational models toward patient-specific prediction, and analyzes the potential roles and practical boundaries of AI methods in model acceleration, input construction, and parameter calibration. Future developments in fracture-healing modeling will not merely aim to improve the computational efficiency of individual simulations, but will increasingly move toward hybrid modeling frameworks that integrate mechanism-based constraints, data-driven learning, and patient-specific inputs, thereby supporting more reliable individualized prediction of fracture healing.</p>