Ultisols with high physio-chemical protection and high acidity prevent soil organic carbon (SOC) from microbial decomposition and accumulate plant-derived recalcitrant C by forming organo-metal complexes. Therefore, the C stabilization pathways in Ultisols may challenge the current common understanding, which is that the major pathway is the stabilization of microbial-derived C to minerals. In addition, organic amendments are expected to affect C stabilization pathways by promoting SOC decomposition and changing SOC composition. However, limited evidence has been found for Ultisols, and the underlying mechanism remains under discussion. C stabilization processes lead to 13C fractionation and changes in SOC composition among aggregate fractions. Therefore, 13C natural abundance and 13C nuclear magnetic resonance spectrometry approaches were applied to investigate 13C fractionation and SOC composition, especially the relative content of recalcitrant C, such as lignin and lipids, and the humification degree represented by the O-alkyl/alkyl ratio, between aggregate size classes. Five fertilization regimes from an 8-year field experiment were selected: no fertilization (CK), inorganic fertilizer (NPK), NPK + straw (NS), NPK + straw and manure (NSM), and NPK + biochar (NC). The bulk soil was separated into three aggregate fractions: >2000 μm, 250–2000 μm, and < 250 μm. In all treatments, 13C presented greater depletion in microaggregates (<250 μm) than macroaggregates (>2000 μm and 250–2000 μm). Regarding the changes in SOC composition, in non-organic-amended treatments (CK and NPK) and NC, recalcitrant C fractions, such as lignin and lipids, were more abundant in < 250 μm aggregates. With the fresh organic material (OM) application (NS and NSM), the lowest O-alkyl/alkyl ratio was found in the > 2000 μm aggregates, suggesting a high humification degree. Summarizing the δ13C and SOC composition results, in the non-organic-amended treatments and NC, the depletion of δ13C in microaggregates coincided with the enrichment of recalcitrant C, which suggests that the depletion of δ13C was primarily controlled by the preferential substrate utilization mechanism. Thus, labile C with heavier δ13C is preferentially decomposed while partially oxidized recalcitrant C with lighter δ13C was adsorbed to the mineral surface and protected in microaggregates. Because the δ13C value decreases with the aggregate size, C flow (modeled in a scheme) primarily occurred from macroaggregates to microaggregates. With fresh OM input (NS and NSM), the C stabilization pathway was primarily governed by the kinetic fractionation mechanism; that is, heavier δ13C remained after microbial utilization, which was reflected by the association of heavier δ13C and a lower O-alkyl/alkyl ratio. With macroaggregates exhibiting 13C enrichment, the C flow was primarily from microaggregates to macroaggregates. These results highlight the importance of the accumulation and protection of plant-derived recalcitrant C in microaggregates in the C stabilization pathway of Ultisols and emphasize that the C stabilizing mechanisms largely depend on organic amendments.