SAME TALES, DIFFERENT STORIES? A COMPARATIVE MULTIMODAL ANALYSIS OF AI-GENERATED FAIRY TALES

Do generative AI platforms produce comparable narratives and images when prompted with the same fairy tales? This study compares episode selection, narrative organization and visual representation across ChatGPT, Gemini and Microsoft Copilot. The corpus comprises fifteen text-image pairs generated in August 2026 using the platforms’ free versions. Each platform received the same instruction to select a scene from each of five fairy tales and narrate it in approximately two hundred words. Each passage was then supplied to the same platform for visualization. The first outputs were retained without regeneration. A comparative qualitative multiple-case design integrates fairy-tale scholarship, narratology, visual social semiotics and intersemiotic relations. Two researchers independently examined both modes and compared their interpretations. Textual analysis addresses episode selection, temporal organization, focalization and narrative functions; visual analysis examines participants, actions, perspective and composition. The findings reveal different episode selections under identical instructions. Images preserve recognizable characters and motifs while selecting, condensing or expanding narrated events. Thematic recognition therefore does not guarantee event correspondence or equivalent representations of conflict. The conclusions concern this situated corpus and do not establish stable platform styles or general performance.