Accepted for/Published in: Journal of Medical Internet Research
Date Submitted: Jul 28, 2025
Date Accepted: Jan 17, 2026
Date Submitted to PubMed: Jan 18, 2026
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
A Study on the Effectiveness of Al-Assisted Medical Health Education Using Voice Cloning and ChatGPT：A Prospective Randomized Controlled Trial
ABSTRACT
Background:
Traditional patient education often lacks personalization and effectiveness. With the rise of artificial intelligence (AI), voice cloning and large language models like ChatGPT present new opportunities for improving medical education[1,2]. However, their re-al-world application and comparative effectiveness remain underexplored[3,4].
Objective:
Objective:
To evaluate the effectiveness of AI-assisted medical education using voice cloning and ChatGPT, and to compare the impact of physician voice cloning versus patient’s own voice cloning on education outcomes in hospitalized patients.
Methods:
We conducted a prospective randomized controlled trial involving 180 hospitalized pa-tients who required medical education. Participants were randomly assigned to one of three groups: (1) traditional education group (n=60), (2) physician voice cloning educa-tion group (n=60), and (3) patient’s own voice cloning education group (n=60). All groups received standardized education content. Primary outcome was education con-tent compliance rate, evaluated using ChatGPT-4 with pre-validated prompts. Second-ary outcomes included knowledge mastery, satisfaction, treatment adherence, quality of life (SF-36), and psychological status (HADS). A dedicated pre-trial validation was performed to ensure the reliability of ChatGPT-based assessments.
Results:
A total of 174 participants completed the trial. Both intervention groups showed signif-icantly higher education compliance rates and satisfaction than the control group (P<.001). The patient’s own voice cloning group outperformed the physician voice cloning group in content retention (92.5% vs 86.7%), satisfaction, adherence, and qual-ity-of-life domains (P<.05). One-month follow-up revealed improved treatment adher-ence and reduced anxiety and depression in the intervention groups, especially in the patient self-voice group. ChatGPT evaluation showed high consistency with expert scoring (Kappa=0.87), indicating good reliability.
Conclusions:
AI-assisted education using voice cloning and ChatGPT significantly improves patient education effectiveness. Education delivered through a patient’s own cloned voice demonstrates superior outcomes compared to physician voice. ChatGPT serves as an efficient and reliable evaluation tool, supporting scalable and personalized education models in healthcare. Clinical Trial: Chinese Clinical Trial Registry, No: ChiCTR2500101882.
Citation
