Patient engagement, acceptability, and preference of artificial intelligence versus human coaching for diabetes prevention

Abstract

Artificial intelligence-powered lifestyle interventions may expand access to behavioral change support, but patient experience remains critical to real-world uptake and effectiveness. In a phase 3 randomized clinical trial of adults with prediabetes and overweight or obesity ( N = 368), we compared a fully automated AI-driven diabetes prevention program (DPP) with a CDC-recognized human-coach-based DPP. Participants assigned to the AI-led DPP initiated the intervention sooner (median 11 vs 26 days; p < 0.001) and demonstrated higher and more evenly distributed engagement over 12 months. Acceptability ratings were consistently higher for the human-led DPP across all domains, with the largest differences in satisfaction ( p < 0.001). Preference responses also favored human coaching, with 46.5% of AI-assigned participants indicating they would have preferred a human coach compared with 31.5% preferring a fully automated program in the human-led group ( p = 0.012). These findings illustrate a trade-off between scalability and perceived quality of experience. Clinical trial registration: ClinicalTrials.gov NCT05056376 (registered September 24, 2021).