Artificial Intelligence and Student Creativity: An Exploratory Study of Students’ Experiences with AI Tools

This study explores how students utilize AI tools in a short visual design task and how these experiences relate to self-reported creativity and creative self-efficacy, using an exploratory mixed-methods approach. The primary aim is to understand students’ perspectives on AI-supported creation, with quantitative measures serving as contextual evidence. Fifty-eight undergraduate students completed a brief design project supported by AI tools and finished pre- and post-task surveys assessing creativity and creative self-efficacy. A subset of 18 students also joined focus group interviews to provide qualitative insights. High-performing students described using AI as a collaborative partner to augment ideas and validate directions while maintaining a sense of control. In contrast, medium- and low-performing students reported a greater reliance on AI outputs, feeling overwhelmed by the abundance of options, and a reduced confidence in their own creative abilities without AI support. Quantitative analyses revealed no detectable change in creativity or creative self-efficacy from pre- to post-intervention, and negligible associations between these self-reported measures and creative performance. Overall, the findings suggest that AI tools alone may not yield short-term gains in perceived creativity or creative self-efficacy; instead, students’ outcomes and experiences appear to vary with how critically and reflectively they engage with AI. The study highlights the need for pedagogical scaffolding and instructional guidance when integrating AI into creative education. • An exploratory mixed-methods study of students’ AI tool use in an authentic short design task. • No detectable pre-post change in creativity or creative self-efficacy. • Expert-rated creative performance was weakly related to self-reported creativity and efficacy. • High performers described using AI for validation and iterative refinement. • Lower performers reported choice overload, AI dependence, and lower confidence, implying scaffolding and reflection.