The size of non-perturbative corrections to high E_T jet production in deep-inelastic scattering is reviewed. Based on predictions from fragmentation models, hadronization corrections for different jet definitions are compared and the model dependence as well as the dependence on model parameters is investigated. To test whether these hadronization corrections can be applied to next-to-leading order (NLO) calculations, jet properties and topologies in different parton cascade models are compared to those in NLO. The size of the uncertainties in estimating the hadronization corrections is compared to the uncertainties of perturbative predictions. It is shown that for the inclusive k_perp ordered jet clustering algorithm the hadronization corrections are smallest and their uncertainties are of the same size as the uncertainties of perturbative NLO predictions.