AI-driven cultural urbanism: A data-integrated model for learning city development in emerging heritage contexts

This study introduces an AI-enhanced framework for designing culturally rooted Learning Cities, demonstrated through Warinchamrap, Thailand—a heritage-rich secondary city overlooked by conventional urban development. Addressing gaps in smart city models that neglect symbolic landscapes, community narratives, and intergenerational learning, this research merges AI modeling with participatory cultural mapping and spatial analytics to create eight location-specific learning nodes.The methodology comprised five phases: (1) integrating multi-source data (archival, oral history, geospatial, social media); (2) applying AI clustering and natural language processing to identify user personas and spatial patterns; (3) employing generative AI (GPT-4, diffusion models) for culturally appropriate design concepts; (4) evaluating designs via a Design Suitability Score (DSS) framework combining AI metrics and stakeholder validation; and (5) refining architectural designs based on community feedback.Initial AI concepts achieved DSS scores averaging 0.866. After prompt refinement, version 2 designs improved to 0.908. Final architectural designs, informed by AI outputs, maintained strong community alignment with a DSS of 0.893. Projects like Songsarn x Rails of Memory and Warin Light Avenue successfully integrated spatial storytelling, cultural heritage, and adaptive learning design. Findings demonstrate that culturally embedded AI can serve as an effective co-designer for inclusive, memory-driven urban learning environments. This research provides a replicable framework for Learning Cities discourse, synthesizing generative AI, local heritage, and spatial intelligence. It offers valuable insights for developing culturally sustainable smart cities, adaptive tourism, and community-driven urban design, establishing AI as a responsive tool when anchored in cultural semantics and participatory engagement.