Sustainable cities are vital for addressing global environmental and social challenges, yet evaluating their diverse sustainability aspects remains complex. Traditional Multi-Criteria Decision-Making (MCDM) methods often suffer from subjectivity, resource intensity, and cognitive burden due to reliance on small expert pools and pairwise comparisons. This study introduces an integrated framework to overcome these limitations by identifying and prioritizing key factors for urban sustainability evaluation. We employ BERTopic, a transformer-based topic modelling technique, to systematically extract 12 relevant factors from the academic literature. Instead of human experts, we leverage a state-of-the-art generative Artificial Intelligence (AI) model (Gemini 2.5 pro) with chain-of-thought reasoning to provide structured evaluations for these factors across different importance clusters. The inherent uncertainty in these AI-generated judgments is modelled using fermatean fuzzy sets. Finally, the factors are prioritized using the soft cluster rectangle method, eliminating the need for pairwise comparisons. Results indicate that pollution control, water management, and social equity are the highest-priority factors, followed by sustainable transportation, urban ecology, population health, and urban resilience. This study presents a more objective, scalable, and efficient data-driven approach to aid policymakers in strategic urban sustainability planning.