Cognitive Language-Learning Strategies, Perceived Risk, and ChatGPT Use Among Croatian University Students: An EFL-Focused Study

Generative artificial intelligence is increasingly integrated into university students’ language-learning practices, yet less is known about how established cognitive language-learning strategies and perceived risk relate to students’ use of ChatGPT ‘(OpenAI, San Francisco, CA, USA; model/version not specified). This study examined cognitive language-learning strategies, perceived risk, general ChatGPT use, and translation-related ChatGPT use among university students in Croatia. The full sample comprised 307 students, of whom 284 were learning English as a foreign language and completed the EFL-specific measures. Data were collected using a paper-and-pencil questionnaire and analyzed using Spearman correlations, age-adjusted analysis, and ordinal logistic regression. Cognitive language learning strategies were not associated with either general or translation-related ChatGPT use. Perceived risk was weakly negatively associated with general ChatGPT use, and this relationship remained statistically significant after age was taken into account. Age was negatively associated with general ChatGPT use and positively associated with perceived risk but was unrelated to translation-related ChatGPT use. Exploratory analysis showed that greater concern about cognitive over-reliance was weakly associated with more frequent general ChatGPT use and greater reported use for translation. The findings indicate that cognitive language learning strategies, perceived risk, and different forms of ChatGPT use represent distinct aspects of students’ engagement with generative AI and should not be treated as interchangeable indicators of AI-supported language learning.