From AI Anxiety to Strategic Regulation: How University Students Transform Generative AI into a Strategic Learning Resource

Synthesis: Eunjeo Kim (2026) examined how university students engage with generative AI as a strategic learning resource in academic writing tasks, focusing on the role of AI anxiety — a central construct of the knowledge base's Anxiety and Stress concept. Using an explanatory sequential mixed-methods design, the study collected survey data and post-task written reflections from 107 university students.

Key Findings

• AI anxiety is productive, not just a barrier: Higher AI anxiety was positively associated with verification and revision behaviors (β=.24, p<.01). Students worried about plagiarism or accuracy were more likely to question every AI sentence, cross-check sources, and revise rather than accept output uncritically.
• Evaluative capacity drives active engagement: Evaluative capacity predicted revision and selective integration (β=.46, p<.001), while ethical awareness strengthened the translation of evaluation into active, responsible AI use (interaction β=.41, p<.01).
• Four regulatory types (N=107): Uncritical Reliance (18.7%), Selective Integration (34.6%), Evaluative Transformation (31.8%), and Strategic Rejection (14.9%). Effective AI use depended not on willingness to use the tool but on the capacity to question outputs, revise selectively, and maintain authorship responsibility. Uncritical reliance risks cognitive offloading, while the other modes reflect active self-regulated engagement.
• AI literacy as regulatory competence: The study frames AI Literacy in higher education less as technology acceptance and more as a form of regulatory competence grounded in evaluative judgment and ethical responsibility.

What this means for practice

• Learners. Read your own discomfort with AI output as a signal to scrutinize, not a reason to avoid the tool: higher AI anxiety predicted more verification and revision (β = .24, p < .01).
• Learners. Invest in evaluative capacity rather than in deciding whether you "like" AI — it was the strongest predictor of active revision and selective uptake (β = .46, p < .001) and is what separates strategic use from over-reliance.
• Learners. Ground your use in ethical awareness: that combination strengthened the translation of evaluation into responsible AI use (β = .41, p < .01).
• Learners. Document how you revised, attributed, or rejected AI suggestions during a task; the reflective protocols in this study are what turn authorship responsibility from a declaration into a habit.
• Learners. Audit which mode you are in — Uncritical Reliance (18.7% of the sample), Selective Integration (34.6%), Evaluative Transformation (31.8%), or Strategic Rejection (14.9%) — and treat only the first as a warning sign, since rejection after evaluation was itself a regulated strategy.

Limitations

• All 107 undergraduates came from a single general education English course at one South Korean university where generative AI was not formally embedded in the curriculum, so transfer to other disciplines, levels, or instructional designs is untested.
• The writing tasks were relatively short, which limits what the study can say about sustained or discipline-specific academic writing.
• The four regulatory types and the behavioral findings rest on self-reported survey scales and post-task written reflections — attitudinal and retrospective accounts — not on observed behavior.
• The paper reports cross-sectional associations from a hierarchical regression, so the β values establish relationships among anxiety, evaluative capacity, ethical awareness, and regulatory engagement, not that any of them causes the others.

Citation

Kim, E. (2026). From AI anxiety to strategic regulation: How university students transform generative AI into a strategic learning resource. Computers and Education: Artificial Intelligence, 10, 100622.