It is well established that land use change (LUC) can impact soil organic carbon (SOC) in tropical regions, but the long-term effects of LUC on soil quality and C cycling remain unclear. Here, we evaluated how LUC affects soil C cycling in the Amazon region using a 100-year observational chronosequence spanning primary forest-to-pasture conversion and subsequent secondary forest succession. We found a surprising increase in topsoil SOC concentrations 60 years following conversion, despite major losses (> 85%) of forest-derived SOC within the first 25 years. Shifts in molecular composition of SOC, identified with diffuse reflectance infrared Fourier transform (DRIFT) spectroscopy, occurred in tandem with a significant decline in permanganate-oxidizable C (POXC) and β-glucosidase activity (per unit SOC), interpreted as a deceleration of soil C cycling after pasture grasses became the dominant source of C inputs to soil. Secondary forest succession caused rapid reversal to conditions observed under primary forest for β-glucosidase activity but not for SOC molecular composition (DRIFT spectroscopy), reflecting a long-lasting effect of LUC on soil C cycling. Our results show that rapid changes in the origin of SOC occur following deforestation with legacy effects on some indicators of C cycling (e.g. enzyme activity) but not others (e.g. molecular composition). This approach offers mechanistic understanding of LUC in the Amazon basin and can be used to help explain conflicting reports on how deforestation impacts SOC in the region.