ABSTRACT Galaxy clusters grow by accreting galaxies from the field and along filaments of the cosmic web. As galaxies are accreted they are affected by their local environment before they enter (pre-processing), and traverse the cluster potential. Observations that aim to constrain pre-processing are challenging to interpret because filaments comprise a heterogeneous range of environments including groups of galaxies embedded within them and backsplash galaxies that contain a record of their previous passage through the cluster. This motivates using modern cosmological simulations to dissect the population of galaxies found in filaments that are feeding clusters, to better understand their history, and aid the interpretation of observations. We use zoom-in simulations from The ThreeHundred project to track haloes through time and identify their environment. We establish a benchmark for galaxies in cluster infall regions that supports the reconstruction of the different modes of pre-processing. We find that up to 45 per cent of all galaxies fall into clusters via filaments (closer than 1 h−1Mpc from the filament spine). 12 per cent of these filament galaxies are long-established members of groups and between 30 and 60 per cent of filament galaxies at R200 are backsplash galaxies. This number depends on the cluster’s dynamical state and sharply drops with distance. Backsplash galaxies return to clusters after deflecting widely from their entry trajectory, especially in relaxed clusters. They do not have a preferential location with respect to filaments and cannot collapse to form filaments. The remaining pristine galaxies (∼30–60 per cent) are environmentally affected by cosmic filaments alone.