Abstract:

Higher education is increasingly shaped by generative artificial intelligence (AI), which is transforming how students learn, develop skills, and prepare for future employment. Among these tools, ChatGPT has become one of the most widely used due to its ability to provide immediate feedback, adaptive explanations, and continuous support across a broad range of academic tasks. As a result, students are increasingly using it for studying, content comprehension, exam preparation, writing support, idea generation, and problem solving. While previous studies have documented the rapid diffusion of generative AI in higher education and highlighted its educational potential, the mechanisms through which AI-supported skill development is connected to academic and career-related outcomes remain insufficiently understood. In particular, limited attention has been paid to the motivational processes that may explain how the perceived development of academic and transferable skills through ChatGPT is translated into students’ perceptions of learning and employability.

Therefore, this study aims to examine the relationships between ChatGPT-supported skill development, self-efficacy, perceived learning outcomes, and perceived employability among higher education students. The empirical analysis is based on data collected through two waves of a large international student survey conducted during the 2023–2024 and 2024–2025 academic years. The first wave included more than 23,000 students from over 100 countries, while the second wave achieved a comparable sample size and similarly broad international coverage, allowing the proposed model to be tested across two independent and globally diverse student samples. The study focuses on students who reported actual use of ChatGPT and applies Structural Equation Modeling (SEM) separately to both waves in order to estimate direct and indirect relationships among the key constructs. In this framework, ChatGPT-supported skill development is examined as a predictor of self-efficacy, learning outcomes, and perceived employability, with particular attention to the mediating role of self-efficacy.

The results reveal a highly stable pattern across both independent samples. ChatGPT-supported skill development emerges as a strong predictor of self-efficacy and perceived learning outcomes in both waves, confirming the importance of AI-supported competence development for students’ academic self-evaluations. Self-efficacy also has a strong positive association with learning outcomes and plays a central mediating role in explaining how AI-supported skills are translated into more positive perceptions of learning. By contrast, the relationships with perceived employability are positive but notably weaker, suggesting that employability depends on a broader set of contextual and personal factors beyond AI-supported learning experiences alone. Overall, the findings indicate that generative AI contributes to student outcomes primarily through motivational and psychological processes rather than through direct skill accumulation alone. The study therefore provides large-scale international evidence on how AI-supported skill development shapes students’ academic and professional self-evaluations and offers important implications for higher education institutions seeking to integrate AI in ways that support learning, confidence, and readiness for increasingly digitalized educational and labour market environments.

Keywords:

ChatGPT, skill development, career outcomes, higher education, structural equation modeling.