Urban livability assessment remains challenging due to its multidimensional nature and the limitations of conventional evaluation methods, which often rely on subjective weighting or a narrow set of indicators. A significant gap exists in the systematic use and comparison of unsupervised machine-learning algorithms for livability classification, particularly in rapidly urbanizing cities of the Global South where spatial inequalities are substantial and data availability is uneven. This study aims to develop and validate a machine-learning clustering framework for spatial urban livability assessment using fourteen GIS-derived indicators covering infrastructure, services, environmental conditions, security, and socio-demographic characteristics. Three clustering algorithms K-Means, K-Means++, and Hierarchical Clustering were applied to classify the 20 districts of Sari City, northern Iran. While all methods revealed a consistent core–periphery pattern in livability, they differed in cluster separation and stability. Hierarchical clustering produced more distinct partitions, whereas K-Means offered more stable and computationally efficient results. The study avoids case-specific enumeration of district rankings and instead emphasizes the methodological insight gained from the comparative analysis. The findings demonstrate that integrating machine-learning clustering with spatial indicators provides a transparent, data-driven, and general framework for evaluating urban livability. The proposed approach can be readily applied to other cities facing similar planning and equity challenges.