University Students’ Use of Generative AI for English Language Learning: AI Readiness, Challenges, and Perceived Impact on Language Skills

Abstract

Generative AI (artificial intelligence) tools are now widely used by university students for language learning, yet empirical evidence from Central and Eastern Europe remains scarce, and a number of studies rely on closed-ended instruments that capture attitude strength but not the reasoning behind it. This mixed-methods study examines Slovak university students’ use of generative AI for English language learning through three research questions addressing AI readiness, perceived challenges, and the impact on language skills development. A purpose-designed 62-item semi-structured questionnaire was administered to 153 students. The instrument combined 59 closed-ended items with three open-ended questions, creating a within-instrument triangulation design in which qualitative responses corroborated, contextualised, and extended the quantitative findings. An interesting tension emerged from the data: students reported high AI readiness and confidence in responsible use, yet expressed genuine concern about academic integrity and the consequences of AI use. Content reliability was the most prominent challenge, with fabricated references receiving the highest agreement score among the challenge items. The open-ended data reinforced this pattern, with content inaccuracy and over-reliance emerging as the most frequently cited problems. A clear split appeared between text-based and oral skills: students perceived strong AI benefits for vocabulary, grammar, reading, and writing, but reported limited or no perceived benefit for speaking, listening, and pronunciation, with listening falling in the Low range. The qualitative data revealed patterns the Likert scale could not detect: a self-reinforcing cycle of over-reliance and diminished critical thinking, and clear divisions on whether institutions should adopt or limit AI. An institutional gap was also documented: students disagreed that their university had provided adequate AI training, and independently identified AI literacy courses as the most frequently mentioned intervention. This study advances e-learning research by providing one of the first empirical analyses in this specific geographical context. It highlights the importance of including open-ended questions in surveys to reveal contradictions that closed-ended questions might miss. Additionally, it offers concrete recommendations for practice: embedding AI literacy in language curricula, teaching source verification alongside prompting skills, and preserving the dedicated speaking and listening practice for which generative AI is not a substitute.