Current guidelines state that education is fundamental to help people with diabetes modify their lifestyle and prevent ill health and early death. However, many people with diabetes are not receiving adequate education. There is a widespread assumption that transferring knowledge will improve health outcomes but there is little empirical support for this assertion. Indeed, knowledge and behaviour are poorly correlated. Knowledge may be a necessary condition but is rarely a sufficient condition for behaviour change. Single interventions, cognitive or behavioural, have had disappointing results, unsurprisingly given the complexity of human behaviour. The most effective interventions are multifaceted and include education, behavioural and psychosocial elements, and target lifestyle change and factors such as self-efficacy and empowerment. We advocate that educational interventions should have multiple components. They should aim to improve patients’ sense of self-efficacy and empowerment, and build attitudes towards diabetes that will support the lifestyle changes needed for successful self-management. These conclusions have implications for future research and clinical practice.