Acousto-optic optical frequency combs can easily produce several hundreds of mutually coherent lines from a single laser, by successive frequency shifts in a loop containing an acousto-optic frequency shifter. They combine many advantages for multi-heterodyne interferometry and dual-comb spectroscopy. In this paper, we propose a model for an intuitive understanding of the performance of acousto-optic optical frequency combs in the steady state. Though relatively simple, the model qualitatively predicts the effect of various experimental parameters on the spectral characteristics of the comb and highlights the primordial role played by the saturation of the gain medium in the loop. The results are validated experimentally, offering a new insight in the performance and optimization of acousto-optic frequency combs.