The influence of different drug transport routes in intestinal drug permeability screening assays was studied. Three experimental models were compared: the small-intestine-like 2/4/A1 cell model, which has a leaky paracellular pathway, the Caco-2 cell model, which has a tighter paracellular pathway, and artificial hexadecane membranes (HDMs), which exclusively model the passive transcellular pathway. The models were investigated regarding their ability to divide passively and actively transported compounds into two permeability classes and to rank compounds according to human intestinal absorption. In silico permeability models based on two-dimensional (2D) and three-dimensional (3D) molecular descriptors were also developed and validated using external test sets. The cell-based models classified 80% of the acceptably absorbed compounds (FA >/= 30%) correctly, compared to 60% correct classifications using the HDM model. The best compound ranking was obtained with 2/4/A1 (r(s) = 0.74; r(s) = 0.95 after removing actively transported outliers). The in silico model based on 2/4/A1 permeability gave results of similar quality to those obtained when using experimental permeability, and it was also better than the experimental HDM model at compound ranking (r(s) = 0.85 and 0.47, respectively). We conclude that the paracellular transport pathway present in the cell models plays a significant role in models used for intestinal permeability screening and that 2/4/A1 in vitro and in silico models are promising alternatives for drug discovery permeability screening.