When a classical biological control program targets multiple congeneric invasive species, comprehensive native range host-use surveys may be effective in accurately predicting which species would most likely be controlled by a candidate agent in its introduced range. At least fifteen species of Pilosella hawkweeds invade North America, where they aggressively out-compete native vegetation in areas of economic and ecological importance, and continue to spread. A promising candidate agent for multiple Pilosella hawkweeds is the leaf-gall wasp Aulacidea pilosellae. Previous surveys revealing inconclusive patterns of ecological host-use across the native range led us to hypothesize that A. pilosellae host species use may follow a rank-preference order. An extensive survey conducted across the native range of A. pilosellae, assessing all potential host species in four distinct geographic regions, found that host species use varied dramatically in different parts of the survey area. The hypothesized pattern of fixed host-preference ranking by the gall wasp within mixed host-species patches was not supported; instead, when multiple host species co-occurred, a frequency-dependent pattern of host-species use among sites was revealed. Specifically, we found that when multiple putative host species co-occurred, the most abundant host species present within mixed species stands was the only species used at a site significantly more often than expected by chance. We further found that only one species was ever used as a host at a given site, regardless of the number of potential host species available. This first known instance of frequency-dependent host-species use reported for an insect weed biological control agent suggests that similar native range assessments for other multiple congeneric host – insect systems may be universally valuable.