Geopolymers are a class of alternative cementitious material, synthesised via alkaline activation of aluminosilicate precursors. The atomistic nature of geopolymer precursors and binders remain largely elusive due to their inherent amorphicity and heterogeneity; nevertheless, pair distribution function analysis is one experimental technique capable of elucidating accurate structural representations of these amorphous materials, when combined with advanced molecular simulation methods. Here, it is shown that, when analysed in isolation, some valuable information can be gained from pair distribution functions of geopolymer precursors and binders. However, when used in conjunction with molecular simulations such as density functional theory and coarse-grained Monte Carlo analysis, there is the potential to generate accurate atomistic representations revealing new information regarding local structural environments. A novel methodology combining real-space refinements and density functional simulations in an iterative manner (DFT-PDF) has been used to generate an accurate structural representation of the geopolymer precursor metakaolin. The structural representation obtained from this technique reveals the existence of III-coordinated aluminium, which exemplifies the power of the DFT-PDF iterative methodology in probing uncommon chemical environments in materials. The potential for density functional theory-based coarse-grained Monte Carlo analysis to elucidate the structure of geopolymer binders and other heterogeneous materials is also discussed.