Beyond Detection: Repurposing AI-Generated Image Datasets for Visual Literacy in Advertising Design Education

Artificial intelligence-generated imagery is fundamentally transforming how visual content is produced within advertising and design education contexts. Whilst scholarly attention has predominantly focused on AI-generated images either as instruments for enhancing creative productivity or as subjects of technical detection methodologies, their pedagogical dimensions within advertising design education—particularly concerning visual literacy and evaluative judgement—remain insufficiently explored. This paper examines the potential of repurposing publicly accessible AI-generated image detection datasets as pedagogical tools for advertising design education. Through qualitative-informed analysis of a comprehensive dataset featuring matched AI-generated and human-created images with their associated textual prompts, this study identifies recurring compositional and stylistic characteristics that differentiate AI-generated advertising imagery. Findings demonstrate that AI-generated images display marked tendencies towards surface-level realism, stylistic consistency, and literal semantic interpretation, frequently yielding visually persuasive outputs that nonetheless obscure conceptual complexity. The alignment between prompts and resultant images reveals systematic privileging of literal prompt components, often sacrificing strategic or contextual sophistication. These characteristics stand in notable contrast to human-created images, which exhibit greater variability and communicative intentionality. By reconceptualising AI-generated image detection datasets as pedagogical resources rather than exclusively technical instruments, this research advances an education-centred perspective within contemporary discourse on AI literacy across creative disciplines, offering insights for curriculum development and critique-oriented pedagogical practices.