This article outlines the concepts of hands-on and hands-off methods in data-driven learning and summarises empirical research on their relative effectiveness. Drawing on recent discussions of prototypical and marginal operationalisations of data-driven learning (DDL), it then problematises the traditional interpretation of the hands-on, hands-off distinction as paper-based versus computer-based. It recognises that the prototypical vision of DDL is the application of the tools and techniques of corpus linguistics to language learning but also that many operationalisations of DDL depart from this core in significant ways. In combination with the observation that interest in hands-off DDL has been driven by a desire to reduce the cognitive demands of DDL activities, it is recommended that the hands-on and hands-off DDL methods distinction is conceptualised as a continuum of more or less hands-on or hands-off operationalisations. Multiple factors in assessing the degree to which an operationalisation is hands-on or hands-off are outlined. The article concludes with a consideration of the potential for large language model artificial intelligence to intersect discussions around hands-on and hands-off methods in DDL.