| Abstract: |
Cultural heritage education increasingly relies on access to diverse digital materials, including archaeological artefacts, museum records, historical maps, and other multimodal resources. These materials are rarely available in formats directly suitable for classroom instruction. This study examines the use of multimodal generative artificial intelligence (AI) to convert such materials into structured teaching tools. Rather than conceptualising AI as an autonomous lesson-generation system, the research proposes a Human-in-the-Loop framework in which educational value emerges from the interplay of source selection, instructional design, multimodal generation, teacher critique, and iterative refinement. This approach is exemplified by a case study of the ceremonial chariot discovered at Civita Giuliana near Pompeii, which demonstrates how a complex archaeological source can be transformed into a pedagogical resource. Additionally, the paper presents a comparative analysis of various types of heritage inputs, including an archaeological object, a museum artefact record, and a historical map. The findings suggest that multimodal generative AI can substantially enhance cultural heritage instruction when integrated within a teacher-guided workflow that ensures historical accuracy, pedagogical coherence, age appropriateness, and source transparency. |