Coevolved mutualisms, such as those between senita cacti, yuccas, and their respective obligate pollinators, benefit both species involved in the interaction. However, in these pollination mutualisms the pollinator's larvae impose a cost on plants through consumption of developing seeds and fruit. The effects of pollinators on benefits and costs are expected to vary with the abundance of pollinators, because large population sizes result in more eggs and larval seed-eaters. Here, we develop the hypothesis that fruit abortion, which is common in yucca, senita, and plants in general, could in some cases have the function of limiting pollinator abundance and, thereby, increasing fruit production. Using a general steady-state model of fruit production and pollinator dynamics, we demonstrate that plants involved in pollinating seed-eater mutualisms can increase their fecundity by randomly aborting fruit. We show that the ecological conditions under which fruit abortion can improve plants fecundity are not unusual. They are best met when the plant is long-lived, the population dynamics of the pollinator are much faster than those of the plant, the loss of one fruit via abortion kills a larva that would have the expectation of destroying more than one fruit through its future egg laying as an adult moth, and the effects of fruit abortion on pollinator abundance are spatially localized. We then use the approach of adaptive dynamics to find conditions under which a fruit abortion strategy based on regulating the pollinator population could feasibly evolve in this type of plant–pollinator interaction.