Following the initial restoration phase of China's grasslands, grassland management transitioned from the long-implemented "grazing prohibition by enclosure" policy to the comprehensive adoption of a "grass-livestock balance" approach. The reason is that herders and local governments cannot afford the cost of the longer grazing prohibition and thus demand grazing on the initially restored grasslands. However, the preliminarily restored grasslands remained ecologically vulnerable. Especially, some slow variables, such as soil nutrients and species composition of plant community, cannot recover during the several years of grazing prohibition. Consequently, there was an urgent need to determine appropriate utilization thresholds-specifically, optimal grazing intensities for grass-livestock equilibrium-to simultaneously achieve sustainable grassland recovery and balanced pastoralism. This study focused on typical preliminarily restored grasslands in the Xilingol region, where we established an aboveground biomass utilization gradient to quantitatively analyze variation patterns of ecological indicators along this gradient and determine appropriate utilization thresholds. During each removal treatment, we first used a drone-carried multispectral camera to observe the Normalized Difference Vegetation Index (NDVI) of the whole experimental site. At the same time, we measured the aboveground biomass in the non-treated part of the experimental site through the harvesting method. Based on the above actually measured data, we established an NDVI-aboveground biomass model with an R² of approximately 0.80. Then, we used the above model and the observed grid NDVI data for all treatment plots to estimate their aboveground biomass in a non-destructive method. Finally, we determined the removal amount of aboveground biomass for each plot based on using a randomized complete block design. Hence, we established an aboveground biomass utilization gradient from 0 to 100% with a 10% interval. The results demonstrated that the appropriate utilization threshold for the preliminarily restored grasslands in this region reached only half of that for native grasslands in the same area. Specifically, while the native grasslands exhibited an appropriate utilization rate of 40% according to the Ministry of Agriculture's industry standards, the degraded grasslands could sustain only a 20% utilization rate. When utilization exceeded this threshold, the preliminarily restored grasslands showed significant declines in aboveground biomass, vegetation height, coverage, and biomass of climax community species, interrupting grassland recovery trends. These findings supported the management principle advocated by the Ministry of Agriculture's industry standards, which recommended reducing utilization rates for moderately degraded grasslands. Therefore, management preliminarily restored grasslands required adoption of lower utilization rates than those for native grasslands. Grassland management authorities needed to establish a dynamic linkage mechanism integrating three critical factors: actual ecological recovery status, appropriate utilization thresholds, and subsidy quotas. This approach ensured sustained grassland recovery even after grazing resumption.