Plants and ants have widespread relationships that are commonly mediated by the offer of extrafloral nectar (EFN) to ants that protect plants against herbivores. However, these ant–plant interactions are highly facultative and vary in time and space, mainly depending on the characteristics of the ant species, such as density and aggressiveness. In general, the outcomes of these relationships are positive, but in some cases, the presence of ants is neutral or negative to plants. Some studies suggest that aggressive attacks or merely the presence of ants might reduce the visitation rate of insect pollinators, such as bees, to flowers. We used experimental manipulation in natural conditions to test the hypothesis that ants on flowers of EFN-bearing plants might be recognized as a danger by pollinators (bees) and reduce the plant fitness (fruit-set). Our results show that the avoidance that ant bodyguard species feeding on EFNs of the Malpighiaceae Heteropterys pteropetala cause in pollinators, is not enough to decrease plant fruit-set. However, ants were indeed identified as a danger to pollinators as hypothesized and as suggested for other plant–pollinator relationships: flowers with plastic ants placed on the petals produced significantly fewer fruits than other treatments (using instead a plastic circle) or the control (natural condition). Indirect costs of facultative mutualisms are the focus of few studies and have been performed only rarely in the Neotropics; our results show that mutualism must be considered in multitrophic interactions studies for a better understanding of the functioning of the system. © 2014 Elsevier GmbH