Hydration of silicic volcanic glass forms perlite, a dusky, porous form of altered glass characterized by abundant “onion-skin” fractures. The timing and temperature of perlite formation are enigmatic and could plausibly occur during eruption, during post-eruptive cooling, or much later at ambient temperatures. To learn more about the origin of natural perlite, and to fingerprint the hydration waters, we investigated perlitic glass from several synglacial and interglacial rhyolitic lavas and tuffs from the Yellowstone volcanic system. Perlitic cores are surrounded by a series of conchoidal cracks that separate 30- to 100-µm-thick slivers, likely formed in response to hydration-induced stress. H₂O and D/H profiles confirm that most D/H exchange happens together with rapid H₂O addition but some smoother D/H variations may suggest separate minor exchange by deuterium atom interdiffusion following hydration. The hydrated rinds (2–3 wt% H₂O) transition rapidly (within 30 µm, or by 1 wt% H₂O per 10 µm) to unhydrated glass cores. This is consistent with quenched “hydration fronts” where H₂O diffusion coefficients are strongly dependent on H₂O concentrations. The chemical, δ¹⁸O, and δD systematics of bulk glass records last equilibrium between ~110 and 60 °C without chemical exchange but with some δ¹⁸O exchange. Similarly, the δ¹⁸O of water extracted from glass by rapid heating suggests that water was added to the glass during cooling at <200 °C. Our observations support fast hydration at temperatures as low as 60 °C; prolonged exposure to high temperature of 175°–225° during water addition is less likely as the glass would lose alkalies and should alter to clays within days. A compilation of low-temperature hydration diffusion coefficients suggests ~2 orders of magnitude higher rates of diffusion at 60–110 °C temperatures, compared with values expected from extrapolation of high-temperature (>400 °C) experimental data. The thick hydration rinds in perlites, measuring hundreds of microns, preserve the original D/H values of hydrating water as a recorder of paleoclimate conditions. Measured δD values in perlitic lavas are −150 to −191 or 20–40 ‰ lower than glass hydrated by modern Yellowstone waters. This suggests that Yellowstone perlites record the low-δD signature of glacial ice. Cooling calculations, combined with the observed high water diffusion coefficients noted for 60–150 °C, suggest that if sufficient hot water or steam is available, any rhyolite flow greater than ~5 m thick can develop the observed ~250-µm hydration rinds within the expected timescale of cooling (weeks–years). As the process of hydration involves shattering of 30- to 100-µm-thick slivers to expose unhydrated rhyolite glass, the time required for hydration may be even shorter. Rapid hydration and formation of relatively thick-walled glass shards allow perlites to provide a snapshot view of the meteoric water (and thus climate) at the time of initial alteration. Perlites retain their initial hydration D/H signal better than thin-walled ash, which in contrast hydrates over many thousands of years with time-averaged precipitation.