Informed consent and shared decision-making have long been weak points in surgical care—patients often leave consultations confused about risks, and surgeons rarely have time to close that gap. A clinical review in a peer-reviewed colorectal surgery journal now maps the growing role of artificial intelligence in restructuring how surgical teams communicate before, during, and after procedures, with implications that extend well beyond the operating room.
The review identifies several active AI applications reshaping perioperative communication. Chatbots deployed before consultations gather medical histories, freeing surgeons to focus on nuanced risk discussions rather than data collection. Large language model–generated decision aids outperformed traditional surgeon-created materials on readability and comprehension measures. Risk-stratification algorithms synthesize multivariable patient datasets to generate individualized outcome projections, moving away from population-level statistics that many patients struggle to contextualize. Notably, the authors highlight that patients with limited English proficiency, multiple comorbidities, or low health literacy stand to gain disproportionately from AI-assisted tools. A separate strand of evidence—studies on ambient AI scribes—documents reduced after-hours documentation burden and increased patient engagement during visits.
This review is confirmatory rather than paradigm-shifting, but it arrives at a meaningful moment. The broader literature on health communication already establishes that decisional conflict—uncertainty patients feel before consenting to surgery—correlates with poorer postoperative satisfaction and adherence. If AI tools genuinely reduce that conflict, downstream health outcomes could follow. The critical caveat is that most cited evidence comes from small pilots and single-institution studies; randomized trials measuring hard outcomes like complication rates or regret are largely absent. Algorithmic bias also looms large: models trained on non-representative datasets risk compounding disparities rather than resolving them. Regulatory frameworks for AI-generated medical communications remain nascent, and liability for AI-assisted consent remains legally unresolved. Treating this evidence base as promising but preliminary is the analytically sound position.