Carbon isotopic signatures recorded in vertebrate tissues derive from ingested food and thus reflect ecologies and ecosystems. For almost two decades, most carbon isotope-based ecological interpretations of extant and extinct herbivorous mammals have used a single diet–bioapatite enrichment value (14). Assuming this single value applies to all herbivorous mammals, from tiny monkeys to giant elephants, it overlooks potential effects of distinct physiological and metabolic processes on carbon fractionation. By analysing a never before assessed herbivorous group spanning a broad range of body masses—sloths—we discovered considerable variation in diet–bioapatite d13C enrichment among mammals. Statistical tests (ordinary least squares, quantile, robust regressions, Akaike information criterion model tests) document independence from phylogeny, and a previously unrecognized strong and significant correlation of d13C enrichment with body mass for all mammalian herbivores. A single-factor body mass model outperforms all other single-factor or more complex combinatorial models evaluated, including for physiological variables (metabolic rate and body temperature proxies), and indicates that body mass alone predicts d13C enrichment. These analyses, spanning more than 5 orders of magnitude of body sizes, yield a size-dependent prediction of isotopic enrichment across Mammalia and for distinct digestive physiologies, permitting reconstruction of foregut versus hindgut fermentation for fossils and refined mean annual palaeoprecipitation estimates based on d13C of mammalian bioapatite.