Human skin consists of several layers with distinct dielectric properties. Physiological processes leading to changes in dielectric properties of the specific layers can potentially be non-invasively monitored employing dielectric spectroscopy. So far no comprehensive skin and underlying tissue model is available for this purpose in the frequency range between 1 and 100 MHz. Focusing on this dispersion-dominated frequency region, different multilayer skin models are investigated. First, with sublayers obtained from two-phase mixtures, second, three-phase mixtures of shelled cell-like ellipsoids and finally, multiphase mixtures obtained from numerical models of single cells generated using a flexible surface parametrization method. All models are numerically evaluated using the finite-element method and a fringing field sensor on the top of the multilayer system serving as a probe. Furthermore, measurements with the sensor probing skin in vivo were performed. The validity of the models was tested by removing the uppermost skin layer, the stratum corneum (SC). It was found that only a three-phase mixture (extracellular medium, cell membrane and cytoplasm) at least can qualitatively reproduce the measured dispersion still occurring without the SC if the model is set up without a priori knowledge of the dispersive behaviour as e.g. a Cole–Cole fit to measured data. Consequently, microstructural features of tissue have to be part of any accurate skin model in the MHz region.