Dairy farming in the Peruvian Amazon region is mainly in the hands of smallholders. Most of them have a low number of cows but low production levels, being unknown their current contribution to the greenhouse gas emissions (GHG) per unit of milk produced. For this reason, the objective of this study was to estimate the carbon footprint (CF) of milk production (in kg of CO₂equivalents (CO₂e) per kg fat and protein corrected milk (FPCM)) on dairy farms of San Martín region, in the Peruvian Amazon. Emissions of GHG were estimated using equations, following the 2019 refinement of the 2006 IPCC Guidelines. A farm characterization and a cradle to farm gate analysis were conducted on eight representative dairy farms using the mass allocation method. Results showed an average milk production of 10 ± 0.82 kg milk/cow/day, a Gyr x Holstein crossbreds as predominant breed, use of cultivated grasses such as Brachiaria brizantha, living fences (Guazuma ulmifolia Lam) as the predominant silvopastoral arrangement, and low level of external inputs such as feed additives or grains. In relation to CF, an average value of 2.17 ± 0.62 kg CO₂e/kg FPCM was obtained, being enteric fermentation the most important source with 1.74 ± 0.51 kg CO₂e/kg FPCM, followed by the manure management (0.24 ± 0.06 kg CO₂e/kg FPCM), land use (0.12 ± 0.03 kg CO₂e/kg FPCM) and energy/transport (0.05 ± 0.03 kg CO₂e/kg FPCM). Differences between farmers were found, obtaining lower CF values (1.63 vs. 3.10 kg CO₂e/kg FPCM) in farms with better feed quality, higher levels of production, and a higher percentage of lactating animals compared with dry cows. We conclude that dairy farms in the Peruvian Amazon region may reduce their current levels of CF if they improve their current feeding practices. However, further studies considering the estimation of carbon sequestration of trees, soil and crops, and the carbon footprint of the off-farm feeds are also necessary to determine the real carbon balance of these systems.