The distribution of the fossil fuel component in atmospheric CO$_2$ cannot be measured directly at a cheap cost. Could anthropogenic tracers with source patterns similar to fossil fuel CO$_2$ then be used for that purpose? Here we present and evaluate a methodology using surrogate tracers, CO, SF$_6$, and C$_2$Cl$_4$, to deduce fossil fuel CO$_2$. A three-dimensional atmospheric chemistry transport model is used to simulate the relationship between each tracer and fossil fuel CO$_2$. In summertime the regression slopes between fossil fuel CO$_2$ and surrogate tracers show large spatial variations for chemically active tracers ( CO and C$_2$Cl$_4$), although C$_2$Cl$_4$ presents less scatter than CO. At two tall tower sites in the United States ( WLEF, Wisconsin, and WITN, North Carolina), we found that in summertime the C$_2$Cl$_4$ (CO) versus fossil CO$_2$ slope is on average up to 15% ( 25%) higher than in winter. We show that for C$_2$Cl$_4$ this seasonal variation is due to OH oxidation. For CO the seasonal variation is due to both chemistry and mixing with nonanthropogenic CO sources. In wintertime the three surrogate tracers SF$_6$, C$_2$Cl$_4$, and CO are about equally as good indicators of the presence of fossil CO$_2$. However, our model strongly underestimates the variability of SF$_6$ at both towers, probably because of unaccounted for emissions. Hence poor knowledge of emission distribution hampers the use of SF$_6$ as a surrogate tracer. From a practical point of view we recommend the use of C$_2$Cl$_4$ as a proxy of fossil CO$_2$. We also recommend the use of tracers to separate fossil CO$_2$. Despite the fact that the uncertainty on the regression slope is on the order of 30%, the tracer approach is likely to have less bias than when letting one model with one inventory emission map calculate the fossil CO$_2$ distribution.