Proceedings of the 70th International Academic Conference, Madrid

FROM AI-CURIOUS TO AI-CONFIDENT: A LONGITUDINAL STUDY OF STUDENT AI ADOPTION AND THE INSTITUTIONAL RESPONSE IN HIGHER EDUCATION, 2024–2026

CLAUDIA VAN DER VORST

Abstract:

The release of generative AI tools such as ChatGPT in late 2022 triggered one of the fastest behavioral shifts ever observed in higher education. Within roughly three years, what began as cautious experimentation among a small group of early adopters has become a routine part of how students learn, research and write. Higher education institutions, by contrast, are widely perceived to be moving more slowly. The present study addresses how large this gap actually is, how it has developed over time, and whether existing institutional responses are substantive or largely symbolic. The analysis draws on a longitudinal student survey conducted at a European university of applied sciences and at its partner institutions abroad. Four AI-focused waves were collected between the summer semester of 2024 and the summer semester of 2026 (n = 207), complemented by an earlier 2023 baseline on teaching preferences (n = 28). In parallel, a cross-sectional dataset of 87 partner universities in more than thirty countries was generated by returning exchange students, who reported on observed institutional practice rather than on institutional self-description. The overarching research question is: How does the rapid rise in AI fluency among higher education students compare with the responses of the institutions where they study, and what does this mean for educators? Three hypotheses are tested. H1 proposes that between 2024 and 2026, students moved from being AI-curious to AI-confident. By 2026, student proficiency evolved toward source-grounded AI and multimodal learning. The strategic use of tools like NotebookLM for source-based grounding and the creation of AI-generated study podcasts marks a transition from generic experimentation to advanced digital autonomy. H2 proposes that a measurable adaptation lag exists between student behavior and institutional response in 2026. H3 proposes that, where institutions have introduced AI policies, this policy presence is not systematically coupled with substantive measures such as licenses or assessment redesign. Across the four waves, self-reported AI experience and preparedness rose markedly before stabilising at a high plateau in 2025–2026. The institutional dataset shows that policy communication is the dominant institutional response, while resourcing and assessment redesign remain rare. The paper concludes that educators, rather than policies, constitute the immediate lever for closing the gap, and develops practical implications for teaching, assessment and shared governance.

Keywords: Genertiv AI, Higher Education, Longitudinal Study, Studient adoption, Institutional response, Policy-practice decoupling, AI literacy

DOI: 10.20472/IAC.2026.070.020

PDF: Download