Unraveling the mechanisms involved in castration- and therapy-resistant prostate cancer has led to a renewed interest in androgen receptor (AR)-targeted therapeutics. Anti-androgens that block the activity of the AR therefore remain a valid therapeutic option. However, they must be more effective than, or display a distinct mechanism of action or binding mode from those of bicalutamide and hydroxyflutamide, which are currently in clinical use. For that reason, the second-generation anti-androgen MDV3100 was developed. MDV3100, however, shares its 4-cyano-3-(trifluoromethyl)phenyl group with bicalutamide and hydroxyflutamide required for binding to the AR. In this work, we used a combined strategy to find new antagonist structures distinct from the 4-cyano-3-(trifluoromethyl)phenyl group to avoid cross-resistance for these compounds and to find structures without agonist activity on mutant ARs (AR W741C and AR T877A). We found two novel chemotypes with AR-antagonistic activity (IC50: 3–6 μM) by virtual screening and confirmed their biological activity in an androgen-responsive reporter assay. The design of our computational approach was validated by the observation of strongly decreased or absence of agonistic activity on the two mutant ARs. Further structural derivatization to optimize the potency of these compounds can render these chemotypes into very promising, alternative AR antagonists for prostate cancer therapy.