Glomalin-related soil protein (GRSP), a glycoprotein produced by arbuscular mycorrhizal fungi (AMF), is broadly distributed in terrestrial soils. Due to its recalcitrant molecular structure, strong binding affinity to soil particles, and resistance to microbial degradation, GRSP is increasingly recognized as a major contributor to the stable soil organic carbon pool. Understanding how environmental factors regulate GRSP production and turnover is crucial for evaluating long-term carbon storage in forest ecosystems, especially under climate change scenarios. Although ecologically important, how GRSP responds to soil warming, especially in relation to different vegetation types, remains unclear and hinders accurate predictions of soil carbon dynamics in subtropical forests. This study examined how soil warming affects GRSP levels in the rhizosphere of two ecologically distinct but regionally significant tree species: Castanopsis carlesii (Hemsl.) Hayata, a dominant species in natural evergreen broadleaf forests, and Cunninghamia lanceolata (Lamb.) Hook, the most widely planted commercial timber tree in southern China. Soil warming was simulated in the field using heating cables to increase the temperature by 4 ℃. Soil samples were collected from seedling plots of both species between June 28 and August 2, 2022, during the peak growing season. These samples were subsequently analyzed for easily extractable GRSP (EE-GRSP), microbial biomass carbon (MBC), and ammonium nitrogen (NH⁺₄-N) to gain insights into microbial and nutrient dynamics under warming. To explore the associations between GRSP and key soil biogeochemical factors, redundancy analysis (RDA) was employed as a multivariate statistical approach. Results showed that under ambient conditions, C. carlesii soils maintained higher moisture and EE-GRSP levels (2.70 μg/g) than C. lanceolata soils (2.03 μg/g). In contrast, when exposed to warming, C. carlesii soils exhibited a more substantial reduction in soil moisture than C. lanceolata, suggesting that C. carlesii may be more sensitive to elevated temperatures and associated hydrological stress. EE-GRSP levels declined significantly in both species under warming, with a sharper drop in C. carlesii (34.9%) than in C. lanceolata (14.8%). This reduction was accompanied by a 28.4% decline in MBC in C. carlesii soils, suggesting that reduced microbial activity under warming may partly explain the drop in GRSP. Moreover, a significant negative correlation between EE-GRSP and NH⁺₄-N was observed in both species, implying that nitrogen availability may constrain GRSP production. The marked decline in MBC in C. carlesii soils further indicates that warming may indirectly reduce GRSP production by lowering microbial activity. Overall, this study provides new insights into how soil warming and tree species identity interact to shape GRSP dynamics, offering empirical evidence to refine evaluations of soil carbon stability and to guide adaptive management of forest carbon sinks in a changing climate.