Induced polarisation (IP) effects may have significant impact on airborne electromagnetic (AEM) data. They lead to dependence of apparent resistivity on the frequency of the signal. The classic approach to modelling IP consists in deriving analytical models of frequency dependent resistivity of each layer of the model. However, the number of parameters for such models grows fast with the number of layers. Hence the problem of numerical inversion becomes intractable due to high dimensionality and ill conditioning. This work suggests an approach to overcoming this problem. We show that the effects of IP are concentrated in relatively small number of layers and propose a simple algorithm for finding them. The results of inverting real data showing strong IP are presented.