Most cases of type 2 diabetes (T2D) can be prevented by adopting a healthy lifestyle, highlighting that lifestyle modifications should be the primary defence against developing T2D. Although accumulating evidence suggests that exercise can be an efficacious therapy for T2D, especially in conjunction with pharmacological interventions, its long-term effectiveness remains controversial owing to significant adherence challenges. In this narrative review, we combine an evolutionary perspective with epidemiological and prospective interventional studies to examine the efficacy versus effectiveness of varying volumes of exercise prescriptions for treating T2D. Commonly prescribed and recommended volumes of moderate-intensity physical activity (150 min/week) have demonstrated low-to-moderate efficacy in improving glycaemic control, reflected by improvements in glycated haemoglobin levels. Higher exercise volumes have been shown to enhance efficacy. While exercise can be moderately efficacious under the optimal circumstances of short-term exercise interventions (≤1 year), there is little evidence of its long-term effectiveness, primarily due to poor adherence. To date, no study has demonstrated long-term adherence to exercise programmes in individuals with T2D (>1 year). From an evolutionary perspective, the finding that exercise interventions are often ineffective over time is unsurprising. Although often overlooked, humans never evolved to exercise. Exercise is a counter-instinctive behaviour that can be difficult to maintain, even in healthy populations and can be especially challenging for individuals who are unfit or have T2D morbidities. We conclude by presenting several considerations informed by evolutionary logic that may be useful for practitioners, policymakers and advocates of exercise as medicine to improve exercise adherence.