Background: The ventilatory anaerobic threshold (VO2@AT) has been used in preoperative risk assessment and rehabilitation for many years. Our aim was to determine the interobserver variability of AT using cardiopulmonary exercise (CPET) data from a large epidemiological study (SHIP, Study of Health in Pomerania). Methods: VO2@AT was determined from CPET of 1,079 cross-sectional volunteers, according to American Heart Association guidelines. VO2@AT determinations were compared between two experienced physicians, between physicians and qualified medical assistants, and between physicians or medical assistants and software-based algorithms. For the first 522 data sets, the two physicians discussed discrepant readings to reach consensus; the remaining data sets were analyzed without consensus discussion. Results: VO2@AT was detectable in 1,056 data sets. The physicians recorded identical VO2@AT values in 319 out of 522 cases before consensus discussion (61.1%; intraclass correlation coefficient [ICC]: 0.90; 95% confidence interval [CI]: 0.88–0.92) and in 700 out of 1,056 cases overall (66.3%; ICC: 0.95; 95% CI: 0.95–0.96), with an interobserver difference of 0 ± 8% (95% limits of agreement [LOA]: ±161 mL/min). The interobserver difference was − 2 ± 18% (95% LOA: ±418 mL/min) between a physician and medical assistants, and − 19 ± 24% to − 22 ± 26% (95% LOAs: ±719–806 mL/min) between physicians or medical assistants and software-based algorithms. Conclusions: Experienced physicians show high agreement when determining AT in asymptomatic volunteers. However, agreement between physicians and qualified medical assistants is lower, and there is substantial deviation in AT determination between physicians or medical assistants and software-based algorithms. This must be considered when using AT as a decision tool.