The neurosteroids, allopreganolone (ALLOP) and tetrahydrodeoxycorticosterone (THDOC), are potent endogenous modulators of mammalian GABA_A receptors. Determining the physiological roles of steroid modulation has been hampered by a lack of antagonists that selectively prevent steroid binding to specific GABA_A receptor isoforms. To circumvent this problem, we have identified the sites of neurosteroid action on GABA_A receptors to allow the generation of transgenic animals where specific GABA_A receptor isoforms are rendered insensitive to neurosteroids. ALLOP and THDOC have two distinct effects on mammalian GABA_A receptor function: potentiation of GABA responses, and in the absence of agonist, they can directly activate the receptor. By contrast, insect GABA receptors are virtually insensitive to ALLOP suggesting that they lack one or more critical components of the steroid binding site(s). After reviewing known steroid binding sites in other proteins, these components are likely to be polar residues that can form hydrogen-bonds with the steroid’s oxygen moieties. By systematically replacing polar residues in the transmembrane (TM) domains of murine α1 and β2 GABA_A receptor subunits with the hydrophobic side chains found in the Drosophila GABA receptor subunit, RDL, we have identified critical determinants of steroid action on GABA_A receptors. Receptors were expressed as α1β2γ2S constructs in HEK293 cells and their pharmacology determined under voltage-clamp at -40 mV, using whole-cell patch-clamp electrophysiology. Residues T237 and Q242 in the first TM domain of the α1 subunit were found to be important for steroid activation and modulation of GABA_A receptors, respectively. From dose-response curve analyses, replacing T237 with other polar side chains supported high potency steroid activation (THDOC EC₅₀: wild-type, 1609 ± 637 nM; αT237S βγ, 1220 ± 383 nM, P>0.05, t test, n = 5), whereas replacement with hydrophobic residues abolished steroid activation (αT237V and αT237I, EC₅₀s > 20 μM, the limit of solubility). All T237 mutants tested had no effect on the potentiating actions of ALLOP or THDOC. Replacement of Q242 with polar side chains supported high potency steroid potentiation of EC₁₀ GABA responses (THDOC EC₅₀: wild-type, 282 ± 47 nM; Q242H: 202 ± 64 nM (P>0.05), Q242S: 1109 ± 58 nM (P<0.05) n = 7) whereas hydrophobic substitutions abolished steroid potentiation at concentrations up to 10 μM. The effects of these mutations were specific to neurosteroids and did not affect the actions of three distinct classes of GABA_A receptor modulators (10 nM-1 μM diazepam; 1–50 μM pentobaritone or 300 nM-100 μM mefenamic acid). These data are consistent with T237 and Q242 contributing to two distinct neurosteroid binding sites on GABA_A receptor α subunits and provide a structural basis from which to generate transgenic animals with neurosteroid-insensitive GABA_A receptors.