Rural vitality reflects the multidimensional state of rural territorial systems across economic, social, cultural, and environmental dimensions. However, township-level comparisons of rural vitality and its associated spatial conditions across development pathways remain limited. This study examines 123 township-level administrative units in six selected counties of the Yangtze River Delta. Based on resource endowments, industrial structure, and functional positioning, the six counties were grouped into agriculture-led, industry-led, and culture-and-tourism-led pathways. A rural vitality evaluation system was constructed under a unified full-sample framework to ensure cross-county comparability. The optimal parameters-based geographical detector (OPGD) and multiple linear regression (MLR) were used to examine spatial differentiation and associated factors. The results show that rural vitality exhibits clear township-level spatial clustering, with high-vitality units generally located near county seats, development zones, central towns, transport corridors, or areas with stronger public-service conditions. The selected pathway groups show different spatial patterns: agriculture-led cases show central agglomeration, industry-led cases show a multi-node pattern, and culture-and-tourism-led cases show central agglomeration together with resource-related nodes. Associated factors also vary across the selected cases: in agriculture-led cases, they include agricultural spatial foundations, land-use organization, accessibility, and functional mix; in industry-led cases, they include built-environment structure, functional organization, transport connectivity, and locational accessibility; and in culture-and-tourism-led cases, they include resource-environmental conditions, accessibility, and service-support capacity. Interaction analysis suggests that combinations of spatial conditions help describe pathway-related association patterns. This study provides comparative evidence for differentiated rural planning under different development contexts.