The rapid increase in atmospheric CO₂ concentrations remains a major driver of global climate change, making the development of high-performance carbon capture materials essential. These materials must be capable of operating efficiently under realistic conditions including humidity and industrial gases. Understanding the influence of water on CO₂ adsorption in zeolite structures is therefore crucial for developing robust carbon capture technologies. In this study, the impact of water on CO₂ adsorption in HEU-type zeolites is examined using a machine-learned interatomic potential (MLIP) trained on high-fidelity ab initio data. MLIPs enable atomistic simulations with near-density functional theory (DFT) accuracy while also providing access to extended timescales and lengths. Radial distribution functions (RDFs) and cumulative coordination numbers of HEU zeolite co-loaded with H₂O and CO₂ show that the structure is preserved, as indicated by the absence of changes in the Si–O and Si–Si peaks. The O–O distribution broadens, reflecting the aggregation of hydrogen-bonded water molecules within the pores. The reduction in C–network correlations indicates that water preferentially occupies the adsorption sites, displacing CO₂ and weakening its interaction with the network. These results highlight competitive adsorption and pore obstruction under humid conditions. Therefore, improving the hydrophobicity of the zeolite surface or pre-drying the gas streams may be necessary to maintain adsorption efficiency.