Androgen receptor (AR) action is a hallmark of prostate cancer (PCa) with androgen deprivation being standard therapy. Yet, resistance arises and aberrant AR signaling promotes disease. We sought compounds that inhibited genes driving cancer but not normal growth and hypothesized that genes with consensus androgen response elements (cAREs) drive proliferation but genes with selective elements (sAREs) promote differentiation. In a high-throughput promoter-dependent drug screen, doxorubicin (dox) exhibited this ability, acting on DNA rather than AR. This dox effect was observed at low doses for multiple AR target genes in multiple PCa cell lines and also occurred in vivo. Transcriptomic analyses revealed that low dox downregulated cell cycle genes while high dox upregulated DNA damage response genes. In chromatin immunoprecipitation (ChIP) assays with low dox, AR binding to sARE-containing enhancers increased, whereas AR was lost from cAREs. Further, ChIP-seq analysis revealed a subset of genes for which AR binding in low dox increased at pre-existing sites that included sites for prostate-specific factors such as FOXA1. AR dependence on cofactors at sAREs may be the basis for differential modulation by dox that preserves expression of genes for survival but not cancer progression. Repurposing of dox may provide unique opportunities for PCa treatment.