The Kessler-10 (K10) questionnaire was developed as a screening instrument for nonspecific psychological distress. Although it was validated using epidemiological samples, it is increasingly being used in clinical settings. We sought to determine whether the factor structure of the K10 as established in epidemiological samples, could be replicated in a sample of treatment seeking adults. We used confirmatory factor analysis to investigate two established models of the factor structure of the K10 in a sample of clinic attendees. Neither the single factor model of Kessler et al., Psychol Med. 2002;32:959–976, nor the multifactorial model of Brooks et al., Psychol Assess. 2006;18:62–70 fitted the data. Respecification of the Kessler et al. model to reduce item redundancy did not result in adequate fit. These findings call into question the construct validity of the K10 in treatment seeking samples, and the suitability of the scale for clinical settings may need to be reconsidered.