Accessory and minor phases are frequently used to infer the composition and intensive parameters of magmas. In granites, the most common ferromagnesian phases are biotite and amphibole, which raises the question as to what compositional domains are present among these phases and how do those domains relate to granite composition. Here we present a characterization of biotite and amphibole compositions from S-, I-, and A-type granites. A database of biotite (1215 data points) and amphibole (525 data points) compositions from previously classified S-, I-, and A-type granites has been compiled. Three characteristics can be used to describe the variations in biotite composition including X_Annite^Bt (fraction of Fe²⁺ in the octahedral site), X_Fe^{VI∗ Bt} (fraction of total iron in the octahedral site), and total aluminum (apfu). Three characteristics can also be used to describe variations in amphibole composition including, NK/CNK [i.e., (Na + K)/(Ca + Na + K)], Fe/(Fe + Mg), and total aluminum. Utilizing, for the first time, a random forest (machine learning) model the three characteristics for biotite and amphibole could discriminate the inferred source region of the granites with 82 and 96% accuracy, respectively. Biotite composition can also be used to broadly characterize f_O2,f/_H2Of_HF, f/_H2Of_HCl, and f/H_Ff_HCl of each granite type at a given temperature and pressure. Thus, biotite and amphibole compositions can be used to further characterize granites; however, biotite and amphibole should not solely be used to infer the source of a granite.