ABSTRACT Purpose: to compare the accuracy of different criteria used to analyze transient evoked otoacoustic emissions in schoolchildren. Methods: an accuracy study, where an audiological assessment (audiometry, logoaudiometry, tympanometry) and transient emissions were performed with 70 schoolchildren, from the first to the fifth grade of a municipal school, in Northeastern Brazil (6-14 years, 9.9 ± 2 years), with four criteria, all with signal-to-noise ratio ≥ 3 dB, being: criterion A, in all frequency bands; B, in three consecutive frequency bands; C, in three of the five non-consecutive frequency bands; D, in 2, 3 and 4 kHz. Sensitivity, specificity, accuracy and predictive values with their respective confidence intervals of 95% were analyzed. Results: criterion A showed higher sensitivity (92.31%, 95% CI: 67-98%) and lower specificity (17.35%, 95% CI: 10-29%); criterion C higher specificity (84.21%, 95% CI: 72-91%) and higher positive predictive value (52.63%; 95% CI: 51.63-54.63). Accuracy was 82.85% (95% CI 78.23-87.47) in criterion C and 70% (95% CI: 65.38-74.62) in criterion B. Conclusion: criterion C, signal-to-noise ratio ≥ 3dB in three non-consecutive frequency bands, showed the best accuracy, being considered the best choice as a criterion for the isolated use of transient emissions as a hearing screening procedure, in schoolchildren.