Abstract / Summary
Background: Artificial intelligence (AI) and large language models (LLMs) may make surgical information easier to access and understand. However, these tools cannot replace the clinician’s role in shared decision-making or the requirements for valid informed consent. Methods: This narrative review follows the surgical patient pathway from diagnosis and explanation to option comparison, shared decision-making, and informed consent. It examines conventional digital tools, LLM-assisted communication, personalized prediction, direct consent studies, patient and professional perspectives and legal, ethical, and governance issues. A formal meta-analysis and study-level risk-of-bias assessment were not performed. Results: AI may support several stages of surgical care. Diagnostic and predictive systems can assist clinicians. LLMs can help explain approved information, prepare patients for consultation, and improve consent materials. Conventional interactive tools have the strongest evidence for better comprehension. Evidence for LLM-assisted consent is still based mainly on simulations and document analyses. Patient-facing use requires source control, privacy protection, and visible clinical oversight. AI cannot establish capacity, voluntariness, or final authorization. We propose a five-level framework in which safeguards increase with clinical influence. Conclusions: AI may support the patient’s path from diagnosis to an informed surgical choice when it uses approved information, serves a defined task, and remains under clinical oversight. Current evidence does not support autonomous AI-mediated consent or AI-directed treatment choice. The aim should be a better-informed, preference-sensitive decision, not an automated signature.