SummarySeveral studies in clinical and non‐clinical populations indicate differences between rationally and empirically derived subscales for the Eating Disorder Examination Questionnaire (EDE‐Q), including samples of patients seeking bariatric surgery. This study aimed to use exploratory structural equation modelling (ESEM) to estimate the factor structure of the EDE‐Q and assess for the additive value of alternative measurement of eating disorder symptoms. Adolescents and adults completed the EDE‐Q and a psychiatric evaluation prior to bariatric surgery. Data from 330 participants were analysed using the original four‐factor and modified three‐factor structure of the EDE‐Q using both confirmatory factor analysis (CFA) and exploratory structural equation modelling (ESEM). Age, ethnicity, and body mass index were examined as covariates in the best fitting model, and model subscales used to generate a predictive model of clinician screened DSM‐5 eating disorder diagnoses for criterion validity. A CFA of the four‐factor EDE‐Q provided poor model fit for a pre‐surgical bariatric population, but the three‐factor EDE‐Q and an ESEM of the four‐factor EDE‐Q provided excellent model fit. The Eating Concern subscale of the four‐factor ESEM model significantly predicted eating disorder diagnosis and was positively correlated with age. Our results suggest the ESEM derived factors of the EDE‐Q offered some improvements to the original empirically derived factor structure, as subscale scores based on the original items and cross‐loading items yielded an adequate prediction of clinician diagnoses.