Although deception of floral pollinators is well known among orchids [1, 2], the majority of animal-pollinated plants secure pollination by nectar rewards. The costs and benefits of nectar production remain poorly understood [3–5]. Here, we developed a crossing design to introgress a low-nectar-volume locus of Petunia integrifolia into the genetic background of P. axillaris. The resulting introgression line resembled P. axillaris but produced only one-third of the nectar volume. When exposed simultaneously to low-nectar and wild-type P. axillaris plants, hawkmoth pollinators reduced their probing duration on low-nectar plants but otherwise did not show any signs of discrimination against these plants. However, reduced probing duration resulted in reduced seed production in the low-nectar plants despite their higher reproductive potential as evidenced by hand pollination. In line with this interpretation, we found a positive correlation between probing duration and seed set, and hawkmoth pollination of low-nectar plants that were manually supplemented with nectar to parental levels yielded seed sets similar to hand pollination. Thus, a simple self-serving pollinator behavior—the adjustment of probing time in response to nectar volume—may select against reducing nectar and protect many plant-pollinator mutualisms against a drift toward parasitism.