Restoration of degraded grasslands through improved management is among the possible sustainable solutions to compensate for anthropogenic soil carbon (C) emissions. While several studies have shown a positive effect of rehabilitation on soil C, the impact on soil CO2 emissions is still uncertain. Therefore, this study aimed at quantifying the impact of grassland rehabilitation on soil CO2 emissions in a degraded grassland, South Africa. Commonly used rehabilitation practices were considered, that is rotational grazing (RG), livestock exclosure with fertilizer application (EF) and annual burning (AB), all being compared with traditional free grazing (FG). A total of 2880 in situ measurements of CO2 emissions were performed over 2.5 years under field conditions simultaneously with aboveground biomass, soil temperature, water content and soil organic C (SOC) to understand the changes in C fluxes. The RG performed the best under degraded grasslands by decreasing net CO2 emissions (per g of C) by 17% compared to FG, while EF increased emissions by 76% and AB had similar emissions to FG. The lower net emission under RG is associated with an increase in SOC stocks by 50% and aboveground biomass by 93%, after three years of implementation. Soil CO2 emissions were correlated positively to aboveground biomass and topsoil temperature (r = 0.91 and 0.60, respectively), implying a high effect of grass cover on soil microclimate and microbial activity. These results suggested RG as a potential cost‐effective nature‐based soil management strategy to increase SOC stocks into degraded grassland. However, long‐term trials replicated in different environments are still required.