Dissertation on the adequacy of the CRETA-R methodology as a future teaching resource and analysis of the relevance of the evaluation conducted by ChatGPT and professors: Can ChatGPT facilitate teaching?

This study explores the adequacy of the CRETA-R framework as a teaching resource and the reliability of AI-based assessment using ChatGPT. A mixed-methods design was applied with 427 university students completing a structured task evaluated by both professors and ChatGPT. Quantitative analysis employed Kendall’s tau-b and the Wilcoxon test to compare scores, while qualitative clustering examined student performance across CRETA-R dimensions. Results showed no significant differences between human and AI evaluations, with high alignment in scoring criteria. However, students struggled with prompt refinement and example integration. The strongest predictors of high performance were the Role, Adjustment, and Refinement components. The findings support the use of AI as a viable tool for structured academic assessment, though outcomes are highly dependent on prompt quality. The study highlights the importance of AI literacy and structured prompt design to ensure consistent and pedagogically meaningful evaluation. © 2025 Elsevier B.V., All rights reserved.