Urban Digital Twins (UDTs) support smart city management through real- time monitoring and simulation of physical urban systems, but they provide limited representation of social aspects including perceptions, behaviors, and lived experiences. This paper addresses this gap through a transition from UDTs to Social Digital Twins (SDTs) by integrating artificial intelligence and social media analytics to integrate social data into UDT environment. Social media data from X, Instagram, and TikTok are collected for a case study in Lille, France, and combined with geospatial datasets, a 3D photorealistic city model, and environmental indicators. An AI-driven SDT architecture based on three coordinated agents is then applied: a data preparation agent to clean and structure social media content, a processing agent to extract stakeholder profiles, activity patterns, and emotional indicators, and a validation agent to assess data reliability. From 2,869 validated posts, the SDT generates spatial maps of citizen activity, sentiment, and perceived urban issues. The results show how social concerns, emotions, and engagement cluster across urban space and time, showing socio- spatial dynamics that are not captured by conventional UDTs. The study demonstrates the analytical value of AI-assisted SDTs for social urban analysis while highlighting the importance of validation, ethics, and governance in future deployments.

Urban Digital Twins (UDTs) support smart city management through real- time monitoring and simulation of physical urban systems, but they provide limited representation of social aspects including perceptions, behaviors, and lived experiences. This paper addresses this gap through a transition from UDTs to Social Digital Twins (SDTs) by integrating artificial intelligence and social media analytics to integrate social data into UDT environment. Social media data from X, Instagram, and TikTok are collected for a case study in Lille, France, and combined with geospatial datasets, a 3D photorealistic city model, and environmental indicators. An AI-driven SDT architecture based on three coordinated agents is then applied: a data preparation agent to clean and structure social media content, a processing agent to extract stakeholder profiles, activity patterns, and emotional indicators, and a validation agent to assess data reliability. From 2,869 validated posts, the SDT generates spatial maps of citizen activity, sentiment, and perceived urban issues. The results show how social concerns, emotions, and engagement cluster across urban space and time, showing socio- spatial dynamics that are not captured by conventional UDTs. The study demonstrates the analytical value of AI-assisted SDTs for social urban analysis while highlighting the importance of validation, ethics, and governance in future deployments.