Ambivalent Sexism Theory (Glick & Fiske, 1996) has revolutionised understanding of sexism and generated a new way of examining sexist attitudes using the Ambivalent Sexism Inventory (ASI). One key goal in sexism research is to compare sexist attitudes across different groups, including people with different genders and sexual identities. Before doing so, researchers must be confident that the construct(s) they are comparing are invariant across groups. Given assumptions of heterosexuality, and the central role of heterosexual interdependence, we expected the ASI would be variant across people with different genders and sexual orientations. We conducted multigroup measurement invariance tests between heterosexual women, heterosexual men, lesbian women, and gay men (total N = 1614). Results indicated that hostile sexism and benevolent sexism emerged as separate, related, forms of sexism across groups (i.e., configural invariance was met), but item loadings and intercepts were not equivalent (i.e., loading and intercept invariance was not met). Accordingly, the ASI is not a suitable measurement tool to compare sexist attitudes across sexual minorities (lesbian women and gay men) and majorities (heterosexual women and men). We discuss implications for the centrality of heterosexual interdependence in ambivalent sexism, practical implications for the use of ASI, and we encourage researchers to develop new scales to assess sexism across diverse gender and sexual identities.