To investigate the accuracy of the age-at-onset criterion in those who meet other DSM-5 criteria for attention-deficit hyperactivity disorder, using a prospective population cohort we compared four different approaches to asking those aged 25 years (n = 138) when their symptoms started. Receiver operating characteristic curves showed variation between the approaches (χ(3) = 8.99, P = 0.03); all four showed low discrimination against symptoms that had been assessed when they were children (area under the curve: 0.57–0.68). Asking adults to recall specific symptoms may be preferable to recalling at what age symptoms started. However, limitations to retrospective recall add to debate on the validity of ADHD age-at-onset assessment.