Epilepsy is a common neurological disorder in which magnetic resonance imaging plays a key role. Diffusion imaging based on the molecular diffusion of water has been widely used clinically and in research for patients with epilepsy. Diffusion tensor imaging (DTI), the most common model, has been used for around two decades. Several parameters can be derived from DTI that are sensitive, but non-specific, to underlying structural changes. DTI assumes a single diffusion process following a Gaussian distribution within each voxel and is thus an overly simplistic representation of tissue microstructure. Several more advanced models of diffusion are now available that may have greater utility in the understanding of the effects of epilepsy on tissue microstructure. In this review, I summarise the principles, applications in epilepsy and future potential of three such techniques. Diffusion kurtosis imaging (DKI) characterises the degree to which diffusion deviates from Gaussian behaviour and gives an idea of the underlying tissue complexity. It has been used in both focal and generalised epilepsy and seems more sensitive than DTI. Multi-compartment models separate the signal from extra- and intra-axonal compartments in each voxel. The Composite Hindered and Restricted Model of Diffusion (CHARMED) can characterise axonal density but has not yet been applied in patients with epilepsy. The Neurite Orientation Dispersion and Density Imaging (NODDI) model can determine the intracellular volume fraction (ICVF) and degree of dispersion of neurite orientation. Preliminary data suggest it may more sensitive than conventional and diffusion imaging in localising focal epilepsy.