The recent discovery of circular RNAs (circRNAs) in serum exosomes suggests a novel and potentially useful tool for noninvasive cancer diagnosis. However, there are currently no substantial studies addressing this topic. RNA-sequencing analysis was performed between three pairs of non-small-cell lung cancer (NSCLC) patients and controls. The diagnostic values of exosomal circRNAs were assessed in a training set, validation set 1, and validation set 2. A total of 46 abnormally expressed circRNAs were identified. Among these, circ_0047921, circ_0056285, and circ_0007761 were found to display significant diagnostic validity for NSCLC. In distinguishing NSCLC cases from healthy controls, thepanel of aforementioned circRNAs exhibited area under the curve (AUC) values of 0.926 (95% CI, 0.895-0.956) in the training set and 0.919 (95% CI,0.877-0.962) in validation set 1. Circ_0047921 could distinguish NSCLC cases from chronic obstructive pulmonary disease controls (AUC,0.890; 95%CI,0.831-0.948), whereas the circ_0056285 and circ_0007761 combination could distinguish NSCLC cases from tuberculosis controls (AUC,0.820; 95% CI,0.739-0.902). For an early NSCLC diagnosis, similar results were observed for these circRNAs in distinguishing early-stage NSCLC cases from healthy controls, chronic obstructive pulmonary disease controls, or tuberculosis controls. Circ_0056285 expression was correlated with clinical stages and lymph node metastasis. The exosomal circRNAs, circ_0047921, circ_0056285, and circ_0007761, are promising biomarkers for NSCLC diagnosis in the Chinese population.