The human pancarcinoma-associated epithelial cell adhesion molecule (EpCAM) (EGP-2, CO17-1A) is a well-known target for carcinoma-directed immunotherapy. Mouse-derived mAbs directed to EpCAM have been used to treat colon carcinoma patients showing well-tolerable toxic side effects but limited antitumor effects. Humanized or fully human anti-EpCAM mAbs may induce stronger antitumor activity, but proved to produce severe pancreatitis upon use in patients. To evaluate treatment-associated effects before a clinical trial, we have generated a transgenic mouse tumor model that expresses human EpCAM similar to carcinoma patients. In this study, we use this model to study the in vivo behavior of two humanized and one mouse-derived anti-EpCAM mAb, i.e., MOC31-hFc, UBS54, and MOC31. The pharmacokinetics and tissue distribution of the fully human mAb UBS54 and the mouse-derived MOC31 were largely the same after injection in tumor-bearing transgenic mice, whereas the molecularly engineered, humanized MOC31-hFc behaved differently. Injection of UBS54 and MOC31 resulted in significant, dose-dependent uptake of mAb in EpCAM-expressing normal and tumor tissues, accompanied by a drop in serum level, whereas injection of MOC31-hFc resulted in uptake in tumor tissue, limited uptake by normal tissues, and slow blood clearance. It is concluded that the EpCAM-transgenic mouse model provides valuable insights into the potential behavior of humanized anti-EpCAM mAbs in patients. mAbs sharing the same epitope and isotype but constructed differently were shown to behave differently in the model, indicating that the design of mAbs is important for eventual success in in vivo application.