A radiation dose survey has been undertaken involving 256 patients to investigate the dosimetric impact of breast tomosynthesis screening by employing different breast densities estimated by the Dance model, 50‐50 breast model, and patient‐specific density software: Volpara. Mean glandular dose (MGD) based on the Dance model provided the most realistic dose estimate with an average difference of −3.3 ± 4.8% from the patient‐specific estimation. Average differences of −8.2 ± 6.5% and −7.3 ± 4.7% were observed for the 50‐50 breast model and console MGD, respectively. We conclude that the Dance model should be used for dose calculations in radiation dose surveys and establishing diagnostic reference levels (DRL).