The three presently known enzymes responsible for arsenic-using bioenergetic processes are arsenite oxidase (Aio), arsenate reductase (Arr) and alternative arsenite oxidase (Arx), all of which are molybdoenzymes from the vast group referred to as the Mo/W-bisPGD enzyme superfamily. Since arsenite is present in substantial amounts in hydrothermal environments, frequently considered as vestiges of primordial biochemistry, arsenite-based bioenergetics has long been predicted to be ancient. Conflicting scenarios, however, have been put forward proposing either Arr/Arx or Aio as operating in the ancestral metabolism. Phylogenetic data argue in favor of Aio whereas biochemical and physiological data led several authors to propose Arx/Arr as the most ancient anaerobic arsenite metabolizing enzymes. Here we combine phylogenetic approaches with physiological and biochemical experiments to demonstrate that the Arx/Arr enzymes could not have been functional in the Archaean geological eon. We propose that Arr reacts with menaquinones to reduce arsenate whereas Arx reacts with ubiquinone to oxidize arsenite, in line with thermodynamic considerations. The distribution of the quinone biosynthesis pathways, however, clearly indicates that the ubiquinone pathway is recent. An updated phylogeny of Arx furthermore reinforces the hypothesis of a recent emergence of this enzyme. We therefore conclude that anaerobic arsenite redox conversion in the Archaean must have been performed in a metabolism involving Aio.