Androgen receptor (AR) is a major therapeutic target that plays pivotal roles in prostate cancer (PCa) and androgen insensitivity syndromes. We previously proposed that compounds recruited to ligand-binding domain (LBD) surfaces could regulate AR activity in hormone-refractory PCa and discovered several surface modulators of AR function. Surprisingly, the most effective compounds bound preferentially to a surface of unknown function [binding function 3 (BF-3)] instead of the coactivator-binding site [activation function 2 (AF-2)]. Different BF-3 mutations have been identified in PCa or androgen insensitivity syndrome patients, and they can strongly affect AR activity. Further, comparison of AR x-ray structures with and without bound ligands at BF-3 and AF-2 showed structural coupling between both pockets. Here, we combine experimental evidence and molecular dynamic simulations to investigate whether BF-3 mutations affect AR LBD function and dynamics possibly via allosteric conversation between surface sites. Our data indicate that AF-2 conformation is indeed closely coupled to BF-3 and provide mechanistic proof of their structural interconnection. BF-3 mutations may function as allosteric elicitors, probably shifting the AR LBD conformational ensemble toward conformations that alter AF-2 propensity to reorganize into subpockets that accommodate N-terminal domain and coactivator peptides. The induced conformation may result in either increased or decreased AR activity. Activating BF-3 mutations also favor the formation of another pocket (BF-4) in the vicinity of AF-2 and BF-3, which we also previously identified as a hot spot for a small compound. We discuss the possibility that BF-3 may be a protein-docking site that binds to the N-terminal domain and corepressors. AR surface sites are attractive pharmacological targets to develop allosteric modulators that might be alternative lead compounds for drug design.