This study explores the uncertainty introduced in global assessments of coastal flood exposure and risk when not accounting for water level attenuation due to land-surface characteristics. We implement a range of plausible water-level attenuation values for characteristic land-cover classes in the flood module of the Dynamic and Integrated Vulnerability Assessment (DIVA) modelling framework and assess the sensitivity of flood exposure and flood risk indicators to differences in attenuation rates. Results show a reduction of up to 44% in area exposure and even larger reductions in population exposure and expected flood damages when considering water level attenuation. The reductions vary by country, reflecting the differences in the physical characteristics of the floodplain as well as in the spatial distribution of people and assets in coastal regions. We find that uncertainties related to not accounting for water attenuation in global assessments of flood risk are of similar magnitude to the uncertainties related to the amount of SLR expected over the 21st century. Despite using simplified assumptions to account for the process of water level attenuation, which depends on numerous factors and their complex interactions, our results strongly suggest that an improved understanding and representation of the temporal and spatial variation of water levels across floodplains is essential for future impact modelling.