© 2018 Objectives: To evaluate whether the prospective association between insomnia and mental illness in the general population remained after controlling for multiple confounders, or whether this represented partly remitted prior mental illness. Design: Cohort study. Setting: Australian general population. Participants: The participants were 10,444 people aged 15 or older in the Household, Income and Labour Dynamics in Australia (HILDA) survey who did not meet K10 criteria for likely mental illness at baseline (2013–14). Measurements: The prospective associations of insomnia (yes/no) at baseline with mental illness (yes/no) approximately 2 years later (2015–16), determined from scores on the K10, were evaluated using logistic regression. These were then adjusted for potential confounders including sociodemographic factors, physical health and health behaviors, and baseline and past mental health. Results: Insomnia at baseline increased risk of mental illness onset at two-year follow up (OR 2.23, 95% CI 1.91–2.59, P < .001). This relationship was attenuated but still significant after adjustment for confounding variables (OR 1.72 95% CI 1.46–2.02). Accounting for reverse causality from prior mental ill health and baseline symptoms reduced this further but the relationship remained (OR 1.30, 95% CI 1.09–1.55, P = .003). This effect appeared more robust among those <65 years of age. Conclusions: Insomnia has a consistent prospective relationship with mental illness at two-year follow-up. Insomnia did not appear to be simply a symptom of past, or baseline subclinical, mental illness. This supports the specific targeting of insomnia symptoms in selective preventive mental health initiatives, particularly among those under 65 years of age.