Evaluation of ChatGPT-generated training programmes for kickboxing
Abstract
Study aim
The aim of this research is to evaluate the expert-perceived quality of ChatGPT-generated training programmes for kickboxing.
Material and methods
Three 8-week training programmes for beginner-level kickboxers were generated using prompts with increasing levels of contextual detail. Eleven experienced kickboxing coaches evaluated the programmes using a structured questionnaire consisting of 12 items related to programme organization, sport-specific preparation, and practical applicability. Each item was rated using a five-point Likert scale. Descriptive statistics were calculated, and differences between the programmes were analysed using the Friedman test with post hoc Wilcoxon signed-rank comparisons. Internal consistency and inter-expert agreement were assessed using Cronbach’s α and Kendall’s coefficient of concordance.
Results
The expert-perceived quality of the programmes improved with increasing prompt specificity. The mean composite expert score increased from 1.33 ± 0.47 for Training Programme 1 (Basic prompt) to 2.25 ± 0.89 for Training Programme 3 (Detailed prompt). Statistically significant differences were observed between all program. The largest improvements were observed in programme organization, training progression, training diversity, and physical conditioning. Internal consistency was high across all evaluations (Cronbach’s α = 0.85–0.92).
Conclusion
The findings indicate that the quality of AI-generated training programmes strongly depends on the specificity of the prompts used to generate them. When detailed contextual information is provided, ChatGPT may serve as a supportive tool for generating preliminary training structures. However, the generated programmes should be regarded as preliminary drafts requiring expert review to ensure methodological accuracy, sport-specific relevance, and safe implementation in coaching practice.
© 2026 Denis Volsky, Georgiy Korobeynikov, Olha Borysova, Inna Holovach, published by University of Physical Education in Warsaw
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.