Artificial intelligence in higher education: student use, perceived benefits, and emerging risks

Introduction Artificial intelligence (AI) tools, including large language models and automated writing systems, are rapidly transforming how students engage with academic work in higher education. While institutional discussions have primarily focused on academic integrity and technological capability, comparatively less empirical research has examined students' perspectives on how AI shapes learning, decision-making, and academic expectations. This study investigates (1) how university students use AI in academic work, (2) how they perceive its benefits and risks, and (3) how institutional policy ambiguity influences their educational experiences. Methods A convergent mixed-methods design was employed using survey responses from 467 undergraduate and graduate students at a large public research university, supplemented by student focus groups and faculty interviews. Quantitative analyses included descriptive statistics and Bayesian hierarchical logistic regression models to estimate the prevalence of AI use and assess variation in perceived dependence, policy clarity, and integrity-related anxiety across student characteristics. Qualitative analyses were conducted to provide contextual insight into how students and faculty interpret AI use within evolving norms of teaching, learning, and assessment in higher education. Results AI use was widespread, with most students reporting at least weekly use for studying, researching topics, and writing-related tasks. Students generally perceived AI as improving efficiency, accessibility, and academic support. However, many also expressed concerns regarding over-reliance, diminished critical engagement, and uncertainty about acceptable academic use. Nearly half of respondents reported an unclear understanding of institutional AI-use policies, and integrity-related anxiety was common even among students who did not report intentional misuse. Bayesian model results indicated that these concerns varied systematically by academic level and self-reported financial comfort, suggesting important equity-related differences in how students experience AI within the same institutional environment. Discussion These findings suggest that AI integration in higher education is not solely a technological issue, but also a pedagogical and governance challenge. By foregrounding student perspectives and integrating quantitative and qualitative evidence, this study highlights the importance of clear policy communication, intentional assessment design, and AI literacy initiatives to support equitable and responsible AI use in higher education.