Many novel composite materials have been recently developed for water treatment applications, with the aim of achieving multifunctional behaviour, e.g. combining adsorption with light-driven remediation. The application of surface complexation models (SCM) is important to understand how adsorption changes as a function of pH, ionic strength and the presence of competitor ions. Component additive (CA) models describe composite sorbents using a combination of single-phase reference materials. However, predictive adsorption modelling using the CA-SCM approach remains unreliable, due to challenges in the quantitative determination of surface composition. In this study, we test the hypothesis that characterisation of the outermost surface using low energy ion scattering (LEIS) improves CA-SCM accuracy. We consider the TiO2/Fe2O3 photocatalyst-sorbents that are increasingly investigated for arsenic remediation. Due to an iron oxide surface coating that was not captured by bulk analysis, LEIS significantly improves the accuracy of our component additive predictions for monolayer surface processes: adsorption of arsenic(V) and surface acidity. We also demonstrate non-component additivity in multilayer arsenic(III) adsorption, due to changes in surface morphology/porosity. Our results demonstrate how surface-sensitive analytical techniques will improve adsorption modelling for the next generation of composite sorbents.