Baled silage can be contaminated with several mycotoxins, which cause health and economic problems for farm animals. We investigated the relationship between the visible mold area and deoxynivalenol (DON) concentration of Italian ryegrass baled silage. We subjected the sheep offered with three types of silage (moldy, nonmoldy, and visually nonmoldy part within the mold silage) to a feeding experiment. We assayed silage degradability in vitro by using the rumen liquid of the sheep before and after the feeding experiments. Results revealed no significant relationships among the visible mold area and the outer, inner, and total DON concentration in the silage. In the feeding experiment, the relative forage intake of sheep for each silage type significantly differed between experimental days. The visiting frequency of sheep for the silage varied between experimental days, but relative frequency for each silage type did not significantly differ between experimental days. The in vitro rumen degradability of the silage remained unchanged after the feeding experiment. The intake amount of the visually nonmoldy part and the nonmoldy silages increased at the latter half of the feeding experiment compared with that of the moldy silage, indicating that sheep could determine the silage quality and gradually shift from contaminated silage to noncontaminated silage. Therefore, their intake amount would be reduced if visually moldy silage was fed even though its toxin concentration was low. Conversely, they would prefer silage without visible mold even though its toxin concentration was possibly high because visible mold in the silage did not reflect its toxin concentration.