AbstractThe trace element composition of titanite reflects the temperature, pressure and bulk‐rock composition from which it crystallized. Two geochemical discriminators are identified by applying a support vector matrix (a machine learning classifier) to a global compilation of titanite trace element data. The compilation comprises more than 7,400 analyses of igneous and metamorphic titanite from a wide range of bulk‐rock compositions. First, igneous and metamorphic titanite are differentiated on the basis of Al/Fe and ∑LREE content. Variation in Th/U aids differentiation in composite settings, such as igneous rocks overprinted by metamorphism. Second, titanite from felsic host rocks is distinguished by low Zr/Y and high Fe content. For titanite from igneous rocks, this effectively discriminates titanite from mafic and felsic rocks. Together, these geochemical discriminators may be used to characterize the crystallization setting and host‐rock of an unknown titanite, a valuable tool with applications including petrochronology and detrital provenance analysis.