BACKGROUND This paper examines the specificity and sensitivity of a breath carbon monoxide (BCO) test and optimum BCO cutoff level for validating self-reported tobacco smoking in Indigenous Australians in Arnhem Land, Northern Territory (NT). METHODS In a sample of 400 people (>/=16 years) interviewed about tobacco use in three communities, both self-reported smoking and BCO data were recorded for 309 study participants. Of these, 249 reported smoking tobacco within the preceding 24 hours, and 60 reported they had never smoked or had not smoked tobacco for >/=6 months. The sample was opportunistically recruited using quotas to reflect age and gender balances in the communities where the combined Indigenous populations comprised 1,104 males and 1,215 females (>/=16 years). Local Indigenous research workers assisted researchers in interviewing participants and facilitating BCO tests using a portable hand-held analyzer. RESULTS A BCO cutoff of >/=7 parts per million (ppm) provided good agreement between self-report and BCO (96.0% sensitivity, 93.3% specificity). An alternative cutoff of >/=5 ppm increased sensitivity from 96.0% to 99.6% with no change in specificity (93.3%). With data for two self-reported nonsmokers who also reported that they smoked cannabis removed from the analysis, specificity increased to 96.6%. CONCLUSIONS In these disadvantaged Indigenous populations, where data describing smoking are few, testing for BCO provides a practical, noninvasive, and immediate method to validate self-reported smoking. In further studies of tobacco smoking in these populations, cannabis use should be considered where self-reported nonsmokers show high BCO.