AbstractSustainable management of groundwater relies on robust estimates of aquifer fluxes. As such, recharge can be estimated from groundwater age using analytical models assuming a homogeneous aquifer. We propose to assess how this assumption affects recharge estimates in fluvial deposits. First, we use CHILD, a landscape evolution model, to generate plausible deposits after tens of thousands of years of river evolution. Then, we use the fractional packing model to compute porosity and permeability of unconsolidated sediments assuming a mixture of a coarse and a fine fraction. Finally, we use PFLOTRAN to simulate groundwater flow and mean age over two thousand years of spatially uniform recharge. This reproduces the boundary conditions underlying Vogel's analytical model to estimate recharge rates. Our results are based on 20,000 realizations from varying inputs: grain sizes, aggradation rate of the river, porosity, recharge, etc. For most models, the mean estimate of the recharge rate remains below an absolute error of 25%. But fluvial heterogeneities can lead to local errors ranging from close to zero to several thousand percent, which means that estimates may vary drastically depending on sampling locations. A sensitivity analysis using the δ‐importance measure and CUSUNORO curves exposes how fluvial processes influence the range and distribution of the error. For instance, a high bank erodibility and small coarse grain size under low aggradation rates favor the reworking of deposits, which increases heterogeneity and leads to high errors. This work paves the way for a more reliable estimation of recharge by considering the sedimentary context.