Currently submitted to: Transfer Hub (manuscript eXchange)
Date Submitted: Feb 20, 2026
Open Peer Review Period: Feb 21, 2026 - Apr 18, 2026
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited.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.
Resistance Training Prescriptions Generated by ChatGPT and Google Gemini: Cross-Sectional Comparative Study Abstract
ABSTRACT
Background:
Large language models (LLMs) have expanded the use of generative AI in exercise prescription, but the quality and safety of these recommendations in real-world practice remain uncertain.
Objective:
To compare resistance training prescriptions generated by ChatGPT (GPT-5.1) and Gemini (Flash 2.5) as evaluated by licensed physical education professionals.
Methods:
We conducted an analytical, quantitative, cross-sectional survey with 25 licensed professionals affiliated with the CREF20 council. Two 12-week resistance training programs for a 35-year-old woman with overweight were generated using a standardized prompt—one by each model—and then blindly labeled as Prescription A and Prescription B. Participants rated each prescription on a 5-point Likert scale across five dimensions (quality, clarity, relevance, safety, and usefulness). Data were analyzed in R using Shapiro–Wilk tests for normality and nonparametric comparisons (Wilcoxon and Mann–Whitney U tests). Prespecified subgroup analyses examined differences by age, professional experience, AI usage, and online coaching practices.
Results:
Across the 25 evaluators, no statistically significant differences were observed between ChatGPT and Gemini for the five rated dimensions (all P>.05). Gemini showed a non-significant trend toward higher perceived safety (P=.064; r≈0.37). Subgroup analyses by age, professional experience, AI usage, and online coaching practices likewise showed no significant differences between model outputs (all P>.05).
Conclusions:
ChatGPT and Gemini generated resistance training prescriptions perceived as moderately good and largely equivalent by licensed professionals. These findings suggest that LLMs may be useful as auxiliary tools for drafting training programs, but they do not yet demonstrate sufficient technical refinement to replace professional expertise, particularly regarding individualization, load progression, and systematic risk management.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.