Grasslands have potential to mitigate against climate change because of their large capacity to store soil organic carbon (SOC). However, the long-term impact of grassland management such as burning, which is still common in many areas of the world, on SOC is still a matter of debate. The objective of this study was to quantify the long-term effects of annual burning on CO2 output from soils and SOC stocks. The study was performed on a 62 years old field trial comparing annual burning (AB) to no burning associated with tree encroachment (NB), and to annual mowing (AM) with all treatments laid out in randomized block design with three replicates per treatment. CO2 emissions from soil were continuously measured over two years and were correlated to soil chemical and physical properties. AB and AM produced 30 and 34% greater CO2 emissions from soil than NB (1.80 ± 0.13 vs. 2.34 ± 0.18 and 2.41 ± 0.17 g C-CO2 m− 2 d− 1 for NB, AB and AM respectively). AB and AM also produced greater CO2 emissions from soil and per gram of soil carbon (1.32 ± 0.1 and 1.35 ± 0.1 mg C-CO2 g C− 1 d− 1, respectively) than NB (1.05 ± 0.07 mg C-CO2 g C− 1 d− 1), which corresponded to significant differences of respectively 26% and 29%. Overall, CO2 emissions from soil (per m2) significantly increased with soil water content (r = 0.72) followed by SOC stocks (r = 0.59), SOC content (r = 0.50), soil bulk density (r = 0.49), soil temperature (r = 0.47), C:N ratio (r = 0.46) and mean weight diameter (r = 0.38). These findings suggest that long-term annual burning increases CO2 output from soils. Additional greenhouse gases emissions from burning itself and alternative grassland management techniques were finally discussed.