The purpose of this study was to investigate the impact of different CBCT imaging monitor units (MUs), reconstruction slice thicknesses, and planning CT slice thicknesses on the positioning accuracy of a megavoltage cone‐beam computed tomography (MV‐CBCT) system in image‐guided radiation therapy (IGRT) in head‐and‐neck patients. The MV‐CBCT system was a Siemens MVision, a commercial system integrated into the Siemens ONCOR linear accelerator. The positioning accuracy of the MV‐CBCT system was determined using an anthropomorphic phantom while varying the MV‐CBCT imaging MU, reconstruction slice thickness, and planning CT slice thickness. A total of 240 CBCT images from six head‐and‐neck patients who underwent intensity‐modulated radiotherapy (IMRT) treatment were acquired and reconstructed using different MV‐CBCT scanning protocols. The interfractional setup errors of the patients were retrospectively analyzed for different imaging MUs, reconstruction slice thicknesses, and planning CT slice thicknesses. Using the anthropomorphic phantom, the largest measured mean deviation component and standard deviation of the MVision in 3D directions were 1.3 and 1.0 mm, respectively, for different CBCT imaging MUs, reconstruction slice thicknesses, and planning CT slice thicknesses. The largest setup group system error (M), system error (∑), and random error (σ) from six head‐and‐neck patients were 0.6, 1.2, and 1.7 mm, respectively. No significant difference was found in the positioning accuracy of the MV‐CBCT system between the 5 and 8 MUs, and between the 1 and 3 mm reconstruction slice thicknesses. A thin planning CT slice thickness may achieve higher positioning precision using the phantom measurement, but no significant difference was found in clinical setup precision between the 1 and 3 mm planning CT slice thicknesses.PACS number: 87.55 ne