Abstract
The study aimed to analyze the relationship and predictive capacity of ChatGPT use with respect to self-regulated learning skills in university students, considering frequency of use, quality of use, and strategic academic use. It was conducted using a basic quantitative approach with an explanatory-predictive scope and a non-experimental, cross-sectional design. A total of 335 university students, representing the accessible population, participated through a population census. A Likert-scale questionnaire was administered, and the data were analyzed using partial least squares structural equation modeling (PLS -SEM) with SmartPLS 4.0. The results showed adequate levels of reliability and validity for the measurement model, with factor loadings between 0.791 and 0.902, composite reliability between 0.903 and 0.941, and AVE values between 0.700 and 0.742. The structural model showed positive and statistically significant relationships between frequency of use (β = 0.184; p < 0.001), quality of use (β = 0.327; p < 0.001), and strategic academic use (β = 0.421; p < 0.001) with self-regulated learning competencies. The model explained 61.8% of the variance (R² = 0.618) and demonstrated predictive relevance (Q² = 0.406). It is concluded that strategic academic use has the strongest association and predictive capacity, followed by quality and frequency of use.