From Tourist Perception to Spatial Design: A Multi-Stage Human–AI Decision-Support Framework for Cultural Tourism Homestays

Cultural tourism homestays increasingly integrate accommodation with local culture, natural environments, and experiential value. However, although tourist-generated online reviews have been widely used to analyze tourist experiences, their systematic translation into spatial design decisions remains limited. This study develops a multi-stage decision-support framework that integrates tourist-generated evidence, expert prioritization, site-specific diagnosis, and human–AI collaborative design for cultural tourism homestays. Based on 51,822 valid online reviews from Yunnan, China, natural language processing (NLP), sentiment analysis, grounded-theory-informed qualitative coding, and the Analytic Hierarchy Process (AHP) were integrated to identify tourist experience needs and prioritize design responses. Five experience dimensions were identified: Spatial Environment Experience (SEE), Landscape Resource Perception (LRP), Service Experience Quality (SEQ), Local Cultural Experience (LCE), and Emotional Experience Perception (EEP). Based on judgments provided by 10 experts, AHP assigned the highest design-response priorities to LRP (0.2844) and LCE (0.2670). These weights represent expert judgments concerning how strongly the tourist-derived experience dimensions should be prioritized in design responses; they do not represent tourists’ own rankings of importance. These priorities were subsequently translated into site-specific spatial strategies through a four-stage mechanism linking tourist needs, site-specific problems, spatial intervention targets, and design language. The framework was applied to a homestay in Shuhe Ancient Town, Lijiang, through a GenAI-assisted, designer-led collaborative design process. Post-occupancy evaluation of 228 actual guests provided descriptive feedback on the implemented environment, with LCE (M = 4.332) and LRP (M = 4.113) receiving the highest mean evaluations. By explicitly distinguishing tourist-derived experience needs from expert-derived design priorities and designer-led spatial decisions, the study establishes a traceable multi-actor decision-support pathway connecting tourist-generated evidence, expert prioritization, contextual translation, GenAI-supported exploration, professional design judgment, and post-occupancy feedback.