Graphene oxide (GO) has emerged as a prominent membrane material due to its potential for precise gas and liquid separations enabled by surface modification and channel optimisation. Here, we present tunable graphitic nanofluidic channels-containing GO (gGO) membranes, characterised by a continuous sp2 hybridised carbon lattice distributed across 35 %–72 % of their structure. This architecture enables exceptional water vapour permeability through a Fickian diffusion mechanism, contrasting the non-Fickian transport typically observed in conventional GO membranes. Our findings demonstrate ultrafast water vapour transport through well-defined hydrophobic nanofluidic channels, achieving enhanced water vapour/N2 selectivity. Utilising scalable, top-down multi-scale simulations based on the inverse Ising method, we elucidate the critical role of graphitic sp2 domains in optimising selective diffusion pathways for water molecules. The resulting gGO-based thin-film composite membranes deliver state-of-the-art water vapour/gas separation applications, showcasing their potential for advanced humidification or dehumidification applications.