Spatial optimization of the built environment has become an important strategy to achieve carbon neutrality. However, most existing models ignore the spatial non-stationarity of driving forces. Therefore, this study combines GWR with an integrated SOM-FCM method for unraveling the local variations of built environment drivers of carbon emissions in downtown Guangzhou. The findings show that: (1) The geographical patterns demonstrate an obvious “south-west high and north-east low” emission trend with a clear core-periphery pattern in which intense emissions tend to be significantly concentrated in the downtown historical area. (2) Influencing factors have strong spatial variability; especially, road density and functional intensity aggravate emissions in the mature core, whereas mass transit and green space act as effective mitigating factors in the high-density southern areas. (3) The SOM-FCM framework synthesized these complex patterns into four distinct “carbon emission mechanism zones”, namely Vitality-Traffic Driven, Morphology Dominant, Location-Sensitive, and Industrial-Logistics, each characterized by unique profiles of built environment determinants. Ultimately, this research contributes a data-driven methodology for transitioning from uniform density regulations to zoning-based precision spatial governance in high-density metropolises.