Struggling to learn or learning to struggle? Student identity, anxiety, and academic integrity in AI-assisted writing

Generative artificial intelligence (AI) is reshaping academic writing and, in doing so, reframing how students understand integrity, authorship, and learning. Drawing on semi-structured focus groups with 58 undergraduates in two courses, this study examines students’ lived experiences of AI–assisted writing through a qualitative, practice–oriented lens. We develop a conceptual model in which perceived AI superiority, concerns about social judgment, and uncertainty about detection tools operate as antecedent conditions generating intellectual anxiety and identity disruption. These socioemotional dynamics shape ethical reasoning and authorship practices. Findings show that students engage with AI not as passive users or efficiency seekers but as active curators. They revise, reject, disclose, and reshape AI–generated text to preserve authorial voice, legitimacy, and ownership of learning. Academic integrity thus appears not as simple rule compliance or misconduct avoidance, but as an intentional, negotiated practice enacted through emotionally inflected judgment and contextual decision–making. Anxiety, self–doubt, and fear of misjudgement function as central mechanisms linking AI use to ethical engagement and student agency. Empirically, the study offers qualitative insight into how emotions, identity, and agency intersect in AI–assisted writing. Conceptually, it advances an integrity–as–practice framework and supports process–oriented assessment, engagement, and policy approaches beyond detection–centred or punitive models.