Background and Aims NicAlert produces semi-quantitative assessments of cotinine levels in saliva or urine for verification of smoking abstinence. This study aimed to assess the accuracy of NicAlert readings against a liquid chromatography tandem mass spectrometry assay in smokers who had passed expired-air carbon monoxide (CO) verification but failed NicAlert verification. Design Comparison of NicAlert readings against readings froma reference assay using liquid chromatography tandem mass spectrometry. Setting Geneva, Switzerland. Participants Self-reported non-smokers (n = 92) who in previous testing had CO of 0–3 parts per million (indicating no recent smoking) and reported not using any nicotine product, but had NicAlert readings ≥ 1 (indicating smoking). Measurements NicAlert produces readings of 0, 1 and 2+, which are reported by the manufacturer to correspond to saliva cotinine concentrations of 0–10 ng/ml (indicating not smoking), 10–30 ng/ml and 30+ ng/ml, respectively. Liquid chromatography tandem mass spectrometry was used as the reference. Findings For 82 participants with a NicAlert reading of 1, only two of the liquid chromatography tandem mass spectrometry values were within the purported range of 10–30 ng/ml; 71 were below 4 ng/ml and half the values were below 0.5 ng/ml. Two of the eight participants with NicAlert readings of 2 had laboratory values within the designated range. Neither of the two participants with NicAlert readings of 3 had a cotinine value within the designated range. Conclusions In people who had passed carbon monoxide verification, NicAlert readings yielded a very high false-positive rate in detecting levels of cotinine indicative of smoking.