PURPOSE: To compare volumetric in-room computed tomography (CT) and kilovoltage (kV) cone-beam CT (CBCT) to planar imaging with respect to their ability to localize fiducial markers (FMs) for radiotherapy of prostate cancer. METHODS AND MATERIALS: Image guidance options from two linear accelerators were compared in terms of identifying the center of gravity (COG) of FMs from the isocenter: a Siemens Primatom, where the couch is rotated 180 degrees from the treatment isocenter to the in-room CT vs. electronic portal imaging (EPI); and a Varian OBI system, where kV CBCT, EPI, and planar kV radiographs were compared. In all, 387 image pairs (CBCT = 133; CT = 254) from 18 patients were analyzed. A clinical tolerance of 3 mm was predefined as the acceptable threshold for agreement. RESULTS: COG location on in-room CT and EPI was in agreement 96.9%, 85.8%, and 89.0% of the time in the left-right (LR), superior-inferior (SI), and anterior-posterior (AP) directions, respectively, vs. 99.2%, 91.7%, and 93.2% for the CBCT and EPI analysis. The CBCT vs. kV radiographs were in agreement 100% (LR), 85.4% (SI), and 88.5% (AP), and EPI vs. kV radiographs were in agreement 100% (LR), 94.6% (SI), and 91.5% (AP) of the time. CONCLUSION: Identification of FMs on volumetric or planar images was found to be not equivalent (+/-3 mm) using either linear accelerator. Intrafraction prostate motion, interpretation of FM location, and spatial properties of images are contributing factors. Although in-room CT has superior image quality, the process of realigning the treatment couch to acquire a CT introduces an error, highlighting the benefits of a single isocentric system.