Enhancing Patients' Informed Consent Through AI: Systematic Review.

BACKGROUND: Informed consent is a cornerstone of medical ethics, ensuring that patients understand the risks, benefits, and alternatives of procedures before making health care decisions. However, challenges such as complex medical language, time constraints, and variations in patient literacy often hinder comprehension. Recent advancements in AI offer new opportunities to improve the informed consent process. OBJECTIVE: This systematic review aims to assess AI's effectiveness in enhancing patient understanding and decision-making during the informed consent process. METHODS: Following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, a comprehensive literature search was conducted in PubMed, Embase, and the Cochrane Library to identify studies published in the past 5 years on AI's role in informed consent. Subsequently, the reference lists of selected papers were manually reviewed to include any additional relevant studies. Descriptive and statistical analyses were conducted to evaluate AI's effectiveness, along with tests of homogeneity to assess the feasibility of a meta-analysis. RESULTS: A total of 33 studies published between 2020 and 2025 were included, categorized into 3 domains: AI-generated patient education (n=18, 54.5%), AI-generated consent documentation (n=10, 30.3%), and AI-assisted consent acquisition (n=5, 15.2%). Large language models demonstrated high accuracy, though readability consistently fell below the recommended eighth-grade level. The best-performing model, Copilot, achieved a Flesch-Kincaid Grade Level of 10.59 (±1.22). AI-generated documents improved Flesch Reading Ease Scores by 44% to 122% and reduced required comprehension grade levels by 10% to 47%, and GPT-4 produced significantly more comprehensive consent forms than both Bard Gemini Advanced and human-written documents (P<.001), with accuracy improving by 47% between GPT-3.5 and GPT-4.0. Among AI-assisted consent acquisition randomized controlled trials, AI-assisted patients demonstrated significantly better comprehension of procedural risks than with physician-led consent (P<.001), improved provider-perceived patient understanding in prevasectomy counseling (8.8±1.0 vs 6.7±2.8; P=.047), shorter consultation times (7.7±2.3 min vs 10.6±3.4 min; P=.05), lower postconsent anxiety in total knee arthroplasty (Hospital Anxiety and Depression Scale-Anxiety subscale: 10.48±3.84 vs 12.75±4.12; P=.04), and greater satisfaction with preoperative education (4.22±0.51 vs 3.43±0.84; P<.001). CONCLUSIONS: AI has the potential to improve the informed consent process; however, further research is needed to address ethical concerns and ensure its effective, patient-centered integration into clinical practice.