It is widely accepted that cell fate determination in the cochlea is tightly controlled by different transcription factors (TFs) that remain to be fully defined. Here, we show that Sox9, initially expressed in the entire sensory epithelium of the cochlea, progressively disappears from differentiating hair cells (HCs) and is finally restricted to supporting cells (SCs). By performing ex vivo electroporation of E13.5–E14.5 cochleae, we demonstrate that maintenance of Sox9 expression in the progenitors committed to HC fate blocks their differentiation, even if co-expressed with Atoh1, a transcription factor necessary and sufficient to form HC. Sox9 inhibits Atoh1 transcriptional activity by upregulating Hey1 and HeyL antagonists, and genetic ablation of these genes induces extra HCs along the cochlea. Although Sox9 suppression from sensory progenitors ex vivo leads to a modest increase in the number of HCs, it is not sufficient in vivo to induce supernumerary HC production in an inducible Sox9 knockout model. Taken together, these data show that Sox9 is downregulated from nascent HCs to allow the unfolding of their differentiation program. This may be critical for future strategies to promote fully mature HC formation in regeneration approaches.