The oxygen reduction reaction (ORR) plays a key role in renewable energy transformation processes. Unfortunately, it is inherently sluggish, which greatly limits its industrial application. Sub-nano-cluster-decorated electrode interfaces are promising candidate ORR electrocatalysts. However, understanding the nature of the active sites on these catalysts under electrocatalytic conditions presents a formidable challenge for both experiment and theory, due to their dynamic fluxional character. Here, we combine global optimization with the electronic Grand Canonical DFT to elucidate the structure and dynamics of subnano Pt n clusters deposited on electrified graphite. We show that, under electrochemical conditions, these clusters exist as statistical ensembles of multiple states, whose fluxionality is greatly affected by the applied potential, electrolyte, and adsorbate coverage. The results reveal the presence of potential-dependent active sites and, hence, reaction energetics.