Selective control over Aβ production via γ-secretase modulators (GSM) is a promising strategy for treating Alzheimer's disease, yet the specific binding sites and mechanism of action of GSMs remain unknown. Using the recent cryo-electron microscopy structures of substrate-bound γ-secretase we used two distinct methods to identify four potential binding sites for pyridopyrazine-1,6-dione GSMs. We demonstrate binding to site 4 formed between PS1-TM2, PS1-TM5 and the APP-C83-TM, with experimental activity data correlating significantly (95% confidence) with our computed binding-affinities for this site. Charged protonated GSMs may display higher affinities because of π-cation interaction with the polar residue Tyr115 of PS1-NTF. Surprisingly, the pIC50 of these compounds is largely described (R2 > 0.4 for all of these) by the molecular size, hydrophobicity, and polarizability. We thus believe that we have identified the primary modulator binding site in γ-secretase for these compounds, as well as strong descriptors of GSM potency. Our results are consistent with the FIST model of γ-secretase action and suggest that GSMs work in two ways: The binding affinity itself contributes stability to the ternary enzyme-modulator-substrate complex (tight grabbing), thus preventing early release of the substrate and increasing trimming to shorter, innocent Aβ peptides. At the same time, drug size, hydrophobicity, and polarizability stabilize the more compact semi-open state over the open PS1 state, to make cleavage more precise and complete.