Cell-type specification is guided by transcription factors (TFs) that control specificity and activity of transcription by binding to regulatory DNA elements, such as enhancers. TFs are modular proteins, consisting of structured DNA-binding domains, which allow binding to TF-specific DNA motifs, and intrinsically disordered regions (IDRs), which often harbor “minimal activation domains” that control the activity of the TF. Both, DNA-binding specificity and transcriptional activity of TFs have been extensively studied using ChIP-sequencing and activation domain screens. However, the mechanisms by which TFs establish specific gene expression programs remain poorly understood, as TF-binding or the presence of a minimal activation domain within a TF IDR do not necessarily correlate with target gene expression. Recent studies suggest that non-linear sequence features of TF IDRs facilitate the formation of transcriptional condensates, contributing to both the activity and specificity of TFs, thus indicating a relationship between these two features. In the following, I provide evidence for an evolutionary trade-off between the activity and specificity in human transcription factors encoded as submaximal dispersion of aromatic residues in their IDRs. I identified multiple human TFs that display significant dispersion of aromatic residues in their IDRs, resembling imperfect prion-like sequences. Mutation of dispersed aromatic residues reduced transcriptional activity, while increasing aromatic dispersion in multiple human TFs enhanced transcriptional activity. Furthermore, sequence optimization by increasing aromatic dispersion enhanced in vitro reprogramming efficiency, promoted liquid-like features of condensates formed in vitro, and led to more promiscuous DNA-binding in cells. Together with recent work on enhancer elements, these results suggest an important evolutionary role of suboptimal features in transcriptional control. I propose that engineering of amino acid features that alter condensation may be a strategy to optimize TF-dependent processes, including cellular reprogramming.