A computed tomography (CT) scanner was installed in the linear accelerator room (Primatom) at Morristown. Since June 2000, we have been providing prostate, lung, and liver cancer patients with fusion of CT and linac radiation treatment. This paper describes our registration methods between planning and treatment CT images, and compares treatment localization by CT versus conventional localization by bony landmarks such as portal imaging. For image registration, we printed out beforehand the beam's eye view of the treatment fields. Prostate tumor volume from each Primatom CT slice was mapped on the printouts, and the necessary isocenter shift relative to the skin marks was deduced. No port film was necessary for our Primatom patients. For ten patients we generated digitally‐reconstructed radiographs (DRRs) with bone contrast from the CT scans, and deduced the required shift as the difference between the DRRs of the Primatom CT versus the planning CT This represented the best observable shift should portal imaging be employed. Shift from bony landmark significantly correlated with the Primatom CT shift. Positioning adjustment based on bony anatomy was generally in the same direction as the CT shift for individual patient, but frequently did not go far enough. Our study confirmed that prostate organ motion relative to the bones has an average length of 4.7 mm (with standard deviation of 2.7 mm), and indicated the superiority of CT versus conventional bony structure (such as portal imaging) localization.PACS number(s): 87.53.Kn, 87.53.–j