This article proposes a two-stage procedure aimed at identifying faking in personality tests. The procedure, which can be considered as an extension and refinement of previous item response theory (IRT)-based proposals, combines the information provided by a structural equation model (SEM) in the first stage with that provided by an IRT-based optimal person-fit procedure in the second. In this article, the procedure is presented in its most basic form, which means that it is intended for a two-group or a two-wave SEM and for binary-response IRT models. Then, its functioning is assessed by means of an empirical study based on an instructed-faking design. The results suggest that the effectiveness of the procedure is higher than that of previous related proposals. The usefulness of the proposal in real settings, its limitations, and further extensions are discussed. (Contains 3 tables and 1 figure.)