As artificial intelligence becomes an everyday presence across education, arts and creative technologies, and cultural heritage, the interaction between users and intelligent systems deserves critical examination. This submission presents a systematic review of 95 case studies, 64 in education, 14 in arts, and 17 in heritage — selected via a PRISMA-guided search and expert screening — to map how generative artificial intelligence is embedded at both the interface and interaction levels. We identify nine interface archetypes (e.g., conversational, adaptive dashboards, immersive environments interfaces), eight interaction patterns (e.g., conversing, collaborating, manipulating), and eight main user experience dimensions as observed in case studies.
Our analysis further categorizes six modality-usage patterns—from text, image, audio, and video up to fully multi-modal workflows and distills four main categories of end-to-end application pipelines. Notably, only two studies were found to articulate design-phase guidelines, and limitations cluster around output quality, ethical risks, and a lack of longitudinal evaluations. We conclude with limitations observed, and future research focused on explainability, participatory design, and sustained field deployments. This synthesis provides a foundation for researchers and practitioners seeking to harness
generative artificial intelligence as a responsive, human-centered collaborator.