From clinical prediction to surgical autonomy: Evolution and clinical validation of artificial intelligence in plastic and reconstructive surgery (1990-2026)

Background: Artificial intelligence (AI) has progressively transitioned from experimental computational models to clinically applicable systems across multiple surgical disciplines. Plastic and reconstructive surgery, characterized by complex anatomy and reliance on visual assessment, represents a particularly suitable field for AI integration. Objective: To evaluate the historical development, current clinical applications, and existing evidence gaps of AI in plastic and reconstructive surgery between 1990 and 2026. Methods: Narrative review with structured search strategy of PubMed, Scopus, and Web of Science was performed. Studies evaluating clinically relevant AI applications were included and categorized by decade and clinical domain. Results: Early implementations focused on neural network–based survival prediction in burn patients, reporting high internal accuracy. Subsequent developments introduced computer-assisted diagnostic systems, burn depth classification tools, and microsurgical training platforms. During the 2010s, machine learning and deep learning models achieved diagnostic accuracies exceeding 90% in facial symmetry analysis and tissue viability assessment. Since 2020, AI integration has expanded to telemedicine, free-flap monitoring, thermographic-guided perforator mapping, and large language model–assisted documentation. However, most studies remain retrospective and single-center. Conclusions: AI currently functions as an adjunctive tool that enhances diagnostic objectivity, operative planning, and educational assessment. Prospective multicenter validation and regulatory standardization are required before widespread clinical integration. KEYWORDS Artificial Intelligence; Plastic Surgery; Deep Learning; Microsurgery; Burns; Augmented Reality; Predictive Models.