Subducting oceanic plates carry a considerable amount of water from the surface down to mantle depths and contribute significantly to the global water cycle. A part of these volatiles stored in the slab is expelled at intermediate depths (70–300 km) where dehydration reactions occur. However, despite the fact that water considerably affects many physical properties of rocks, not much is known about the fluid flow path and the interaction with the rocks through which volatiles flow in the slab interior during its dehydration. We performed thermomechanical models (coupled with a petrological database and with incompressible aqueous fluid flow) of a dynamically subducting and dehydrating oceanic plate. Results show that, during slab dehydration, unbending stresses drive part of the released fluids into the cold core of the plate toward a level of strong tectonic under-pressure and neutral (slab-normal) pressure gradients. Fluids progressively accumulate and percolate updip along such a layer forming, together with the upper hydrated layer near the top of the slab, a Double Hydrated Zone (DHZ) where intermediate-depth seismicity could be triggered. The location and predicted mechanics of the DHZ would be consistent with seismological observations regarding Double Seismic Zones (DSZs) found in most subduction zones and suggests that hydrofracturing could be the trigger mechanism for observed intermediate-depth seismicity. In the light of our results, the lower plane of the DSZ is more likely to reflect a layer of upward percolating fluid than a level of mantle dehydration. In our models, a 20–30 km thick DSZ forms in relatively old oceanic plates without requiring an extremely deep slab hydration prior to subduction. The redistribution of fluids into the slab interior during slab unbending also has important implications for slab weakening and the deep water cycle. We estimate that, over the whole of Earth's history, a volume of water equivalent to around one to two oceans can be stored in nominally anhydrous minerals of the oceanic lithosphere and transported to the transition zone by this mechanism, suggesting that mantle regassing could have been efficient even without invoking the formation of high pressure hydrous minerals.