SummaryEnd‐tidal monitors for measuring carbon dioxide (CO2) have become widely available for clinical use in the last two decades. This non‐invasive technology has been previously evaluated in anesthetized veterinary patients, but its accuracy has not been assessed in critical patients. We investigated the usefulness and limits of end‐tidal CO2monitoring in two populations of critical small animal patients: spontaneously breathing dogs and mechanically ventilated patients with healthy and damaged lungs. In analyzing samples from 43 spontaneously breathing dogs and 34 ventilated patients (28 dogs and six cats), the end‐tidal CO2was generally lower than pCO2. The predictive value for hypoventilation was excellent in both populations (100%). The linear correlation of the end‐tidal CO2and arterial pCO2in non‐panting dogs with healthy lungs was 0.84 (p<0.0001), and the 95% confidence interval (CI) of the difference was ± 3.2 mm Hg. However, the measures were uncorrelated in panting dogs (r=0.37, p=0.27), and the 95% CI was ± 13.37 mm Hg. Furthermore, where multiple samples could be obtained in individual patients, the r values and differences of end‐tidal compared to arterial pCO2varied unpredictably. These variations did not appear to be predicted by patient factors such as lung disease. We conclude that the end‐tidal CO2monitor is clinically useful for detecting hypoventilation and monitoring apnea, but it should be supplemented with arterial pCO2determinations if it is important to obtain accu