The purpose of this study is to conduct a psychometric investigation of the Attachment to God Inventory (AGI; Beck & McDonald, 2004) using a national sample of socioeconomically and religiously diverse young adults commonly referred to as millennials (i.e., persons born between 1980 and 1996; Pew Research Center, 2018). Confirmatory factor analytic (CFA) results failed to yield satisfactory fit for the AGI model using the entire sample and a Christian-only subsample. Alternative model specifications that accounted for method factors, higher-order factors, and bi-factors also failed. Exploratory factor analysis (EFA) suggested alternative factor solutions that were cross-validated using CFA. Support for an orthogonal, 2-factor, 8-item model possessed excellent model fit (χ²(20) = 172.186; RMSEA = .051 [.044–.058]; CFI/TLI = .955/.993). Configural, metric, and scale measurement invariance were supported based on gender- and ethnic-identity considerations; however, invariance was not supported based upon religious affiliation. The resulting model consisted of two constructs that were labeled divine rejection (McDonald’s ω = .838 [95% CI: .827–.849]) and divine dependence (McDonald’s ω = .862 [95% CI: .852–.872]) and were found to be invariant only for individuals who identified as Christian. Latent profile analysis (LPA) of the standardized scores of the two factors yielded a five-class solution whose classes were labeled intrinsic, independent, everyday, strained, and detached believers. Class membership was found to be most associated with divine rejection. Rather than considering the resulting model to be an improved model based on attachment to God theory, this study concludes that the resulting model is most consistent with social comparison theory. Several hypotheses and recommendations for future research are made.