Mentoring has recently taken centre stage as one of the primary criminal justice 'interventions' to reduce reoffending, having grown in popularity over the past fifteen years. Its rapid growth has been driven by claims of success within and outwith the criminal justice system, leading some to argue that it has been perceived as a silver bullet (Newburn and Shiner, 2005). This article challenges such claims on three fronts: first, mentoring is an ill-defined concept with weak theoretical foundations; second, the evidence base upon which claims of success are made is limited; and third, transferring mentoring into the coercive and punitive environment of the criminal justice system results in a departure from the very principles and values which are the basis of its usefulness elsewhere. The article utilises the findings from three empirical criminal justice research projects to question claims of widespread and effective mentoring activity with defendants and offenders, suggesting instead that 'interventions' described as mentoring serve as a vehicle to extend the reach of the criminal justice system. At the end of the article we suggest that desistance theory, specifically the Good Lives Model, provides a conceptual framework for taking mentoring in criminal justice forward.