OBJECTIVE To present our single-institution experience with image-guided radiotherapy comparing fiducial markers and cone-beam computed tomography (CBCT) for daily localization of prostate cancer. METHODS From April 2007 to October 2008, 36 patients with prostate cancer received intensity-modulated radiotherapy with daily localization by use of implanted fiducials. Orthogonal kilovoltage (kV) portal imaging preceded all 1244 treatments. Cone-beam computed tomography images were also obtained before 286 treatments (23%). Shifts in the anterior-posterior (AP), superior-inferior (SI), and left-right (LR) dimensions were made from kV fiducial imaging. Cone-beam computed tomography shifts based on soft tissues were recorded. Shifts were compared by use of Bland-Altman limits of agreement. Mean and standard deviation of absolute differences were also compared. A difference of 5 mm or less was acceptable. Subsets including start date, body mass index, and prostate size were analyzed. RESULTS Of 286 treatments, 81 (28%) resulted in a greater than 5.0-mm difference in one or more dimensions. Mean differences in the AP, SI, and LR dimensions were 3.4 ± 2.6 mm, 3.1 ± 2.7 mm, and 1.3 ± 1.6 mm, respectively. Most deviations occurred in the posterior (fiducials, 78%; CBCT, 59%), superior (79%, 61%), and left (57%, 63%) directions. Bland-Altman 95% confidence intervals were -4.0 to 9.3 mm for AP, -9.0 to 5.3 mm for SI, and -4.1 to 3.9 mm for LR. The percentages of shift agreements within ±5 mm were 72.4% for AP, 72.7% for SI, and 97.2% for LR. Correlation between imaging techniques was not altered by time, body mass index, or prostate size. CONCLUSIONS Cone-beam computed tomography and kV fiducial imaging are similar; however, more than one-fourth of CBCT and kV shifts differed enough to affect target coverage. This was even more pronounced with smaller margins (3 mm). Fiducial imaging requires less daily physician input, is less time-consuming, and is our preferred method for prostate image-guided radiotherapy.