Mapping circular city indicators onto the 10R framework: A multi-method AI-powered approach for gap analysis and semantic clustering

: The shift toward circular cities has emerged as a critical pathway for sustainable urban development, yet a unified framework for identifying and classifying circular city indicators across the complete 10R circular economy hierarchy remains underdeveloped. This study systematically maps 241 macro level indicators extracted from scientific literature and policy reports, onto the 10R strategies. To reflect their relative circularity, the 10R strategies are grouped into three maturity tiers: first, smart use and production (Refuse, Rethink, Reduce); second, extend product lifetimes (Reuse, Repair, Refurbish, Remanufacture, Repurpose); and third, focus on material recovery. (Recycle, Recover). A multi-method AI-powered approach was employed, combining pre-trained transformer-based sentence embeddings (MiniLM, MPNet, DeBERTa-v5, SciBERT) with cosine-based semantic similarity measures, ensemble weighting, and clustering techniques to classify indicators across the 10R framework. Semantic mapping reveals a pronounced imbalance: 58% of 241 indicators align with downstream strategies (Recycle, Recover), approximately 31% with upstream preventive strategies (Refuse, Rethink, Reduce), and less than 8% mid-life product life-extension strategies (Reuse, Repair, Refurbish, Remanufacture, Repurpose). Despite comprehensive R-strategy inclusion in the analytical framework, notable conceptual gaps persist—particularly in early-stage, transformative upstream strategies and mid-life interventions. This AI-powered semantic audit provides the first quantitative diagnostic framework for identifying measurement blind spots, enabling policymakers to recalibrate urban CE assessment toward more balanced, system-level circularity transitions.