For computational expediency, regional and global land-surface models (LSMs), especially those coupled to climate simulations, adopt simple algorithms when calculating radiative transfer (RT) and canopy photosynthesis at the vegetated land surface. Nevertheless, the interaction of sunlight with vegetation is recognized as one of the most critical processes represented in LSMs. The present study calculates global, terrestrial Gross Primary Product (GPP) with a version of the land-surface model JULES which has been modified to take explicit account of sunfleck penetration and leaf orientation within the canopy. A comparison with equivalent simulations adopting the Big Leaf (BL) or two-stream (2ST) RT scheme, indicate that current regional/global LSMs may overestimate GPP by 10% globally and up to 25% regionally. Specifically, their use of average light profiles, and consequent neglect of the dispersion in leaf irradiance, at any given height in the canopy leads to both a general overestimation of canopy light-use efficiency (LUE) and a failure to capture the LUE-enhancement under diffuse sunlight (“diffuse fertilization effect”). We also examine the current limitations of regionally/globally implemented RT schemes with respect to canopy architecture. This is done by coupling JULES to the ray-tracing numerical model FLIGHT, the latter simulating light transfer and photosynthesis in both uniform one-dimensional (1-D) canopies and 3-D tree crowns. When the distribution of leaf nitrogen (N) is configured in a manner consistent with field measurements, output from the 3-D and 1-D FLIGHT simulations is fairly similar (predicted GPP differs by ≤5%). Similarly, both Leaf Angle Distribution (LAD), when restricted to its observed range, and leaf-clumping appear to have a minor influence over canopy productivity. We conclude that current LSMs can radically improve their calculation of regional/global GPP by adopting a multilayer approach. This will allow the separate treatment of sunlit and shaded foliage, at discrete heights within the canopy, as well as the accurate representation of active leaf-N.