A rapid diagnosis of pediatric pulmonary tuberculosis (PTB) using Xpert MTB/RIF (Mycobacterium tuberculosis/rifampicin) automated testing on induced sputum (IS) is possible, but the capacity for performing IS is limited. The diagnosis using a nasopharyngeal aspirate (NPA), which can be non-invasively obtained, is desirable. Paired specimens (NPA and IS) were tested using smear, liquid culture and Xpert. The diagnostic accuracy of Xpert and smear was compared with culture for different specimens in children with suspected PTB. There were 535 children [median age 19 months, 117 (21·9%) HIV-infected] who had one IS and one NPA specimen; 396 had two paired specimens. A positive smear, Xpert test or culture occurred in 30 (5.6%), 81 (15.1%) and 87 children (16.3%), respectively. The culture yield was higher from IS (84/87, 96.6%) vs NPA (61/87, 70.1%, P < .001). Amongst children with two paired specimens, 63 culture-confirmed cases occurred [60 (95.2%) IS vs 48 (76.2%) NPA, P = .002]. The sensitivity of two Xpert tests was similar for IS and NPAs [(45/63) 71% vs (41/63) 65%, P = .444)]; the sensitivity of smear was lower for IS (21/63, 33%) and NPA (16/63, 25%). The incremental yield from a second IS was 9 cases (17.6%) by culture and 9 (25%) by Xpert testing; a second NPA increased the culture yield by 10 (26.3%) and Xpert by 11 (36.7%). Xpert specificity was 99.1% (98.1-100) for IS and 98.2% (96.8-99.6) for NPAs. Xpert testing provided faster results than culture (median 0 vs 15 days, P < .001). Xpert testing on 2 NPAs is useful in children with suspected PTB, particularly in settings where IS and culture are not feasible.