Fossil energy analysis, also called cumulative energy demand, is a current method for environmental assessment of agricultural systems. It consists of assessing fossil energy directly consumed on the farm and indirectly consumed to produce farm inputs. Application of the method is based on the sum of the quantity of each farm input multiplied by its energy coefficient, which represents fossil energy use during the life cycle of the input, including extraction, manufacturing and disposal of the raw and auxiliary materials. The range of values for energy coefficients proposed in the literature and variability in their context of application challenges the confidence in estimates of fossil energy use. To offer an initial representation of this uncertainty and to propose improvements for including uncertainty in energy coefficients, a fossil energy analysis completed by uncertainty and sensitivity analyses was applied to dairy farms from two differing French territories: Poitou-Charentes (PC) and Reunion Island (RI). Monte-Carlo simulation of 30,000 sets of coefficients randomly drawn from the ranges of values found in the literature shows a 95% confidence interval ranging from 3.6 to 5.0 MJ L-1 of raw milk for PC and 5.8–8.2 MJ L-1 of raw milk for RI. This represents a variation around the mean of ±16% for PC and ±17% for RI. This also highlights the need for a standardized and clearly defined method for calculating energy coefficients, especially regionally adapted ones, to accurately estimate fossil energy use of agricultural systems throughout the world. Calculation of a more accurate energy coefficient could be associated with a preliminary sensitivity analysis performed with minimum and maximum values of energy coefficients from the literature to identify the most influential coefficients and adapt them to local conditions. This avoids time-consuming adaptation of all energy coefficients, but requires accurate information about life-cycle inputs. Since energy coef