To determine the accuracy and precision of constant infusion transpulmonary thermodilution cardiac output (CITT‐Q) assessment during exercise in humans, using indocyanine green (ICG) dilution and bolus transpulmonary thermodilution (BTD) as reference methods, cardiac output (Q) was determined at rest and during incremental one‐ and two‐legged pedaling on a cycle ergometer, and combined arm cranking with leg pedaling to exhaustion in 15 healthy men. Continuous infusions of iced saline in the femoral vein (n = 41) or simultaneously in the femoral and axillary (n = 66) veins with determination of temperature in the femoral artery were used for CITT‐Q assessment. CITT‐Q was linearly related to ICG‐Q (r = 0.82, CITT‐Q = 0.876 × ICG‐Q + 3.638, P < 0.001; limits of agreement ranging from −1.43 to 3.07 L/min) and BTD‐Q (r = 0.91, CITT‐Q = 0.822 × BTD + 4.481 L/min, P < 0.001; limits of agreement ranging from −1.01 to 2.63 L/min). Compared with ICG‐Q and BTD‐Q, CITT‐Q overestimated cardiac output by 1.6 L/min (≈ 10% of the mean ICG and BTD‐Q values, P < 0.05). For Q between 20 and 28 L/min, we estimated an overestimation < 5%. The coefficient of variation of 23 repeated CITT‐Q measurements was 6.0% (CI: 6.1–11.1%). In conclusion, cardiac output can be precisely and accurately determined with constant infusion transpulmonary thermodilution in exercising humans.