We sought to determine the level of agreement and the reproducibility of two-dimensional (2D) transthoracic (2D-TTE), 2D transesophageal (2D-TEE) and real-time three-dimensional (3D) transesophageal echocardiography (RT3D-TEE) for measurement of aortic annulus size in patients referred for transcatheter aortic valve implantation (TAVI). Accurate preoperative assessment of the dimensions of the aortic annulus is critical for patient selection and successful implantation in those undergoing TAVI for severe aortic stenosis (AS). Annulus size was measured using 2D-TTE, 2D-TEE and RT3D-TEE in 105 patients with severe AS referred for TAVI. Agreement between echocardiographic methods and interobserver variability was assessed using the Bland-Altman method and regression analysis, respectively. The mean aortic annuli were 21,7 ± 3 mm measured with 2D-TTE, 22,6 ± 2,8 mm with 2D-TEE and 22,3 ± 2,9 mm with RT3D-TEE. The results showed a small but significant mean difference and a strong correlation between the three measurement techniques (2D-TTE vs. 2D-TEE mean difference 0,84 ± 1,85 mm, r = 0,8, p < 0,0001; 2D-TEE vs. 3D-TEE 0,27 ± 1,14 mm, r = 0,91, p < 0,02; 2D-TTE vs. 3D-TEE 0,58 ± 2,21 mm, r = 0,72, p = 0,02); however, differences between measurements amounted up to 6,1 mm. Interobserver variability for 2D-TTE and 2D-TEE was substantially higher compared with RT3D-TEE. We found significant differences in the dimensions of the aortic annulus measured by 2D-TTE, 2D-TEE and RT3D-TEE. Thus, in patients referred for TAVI, the echocardiographic method used may have an impact on TAVI strategy.