Internet regulation and 'online harms' are matters of much political and regulatory attention. This debate is beset by issues, including defining 'online harms', respecting freedom of expression, and others. While much of this debate has focused on content hosted by online platforms, comparatively little attention has been paid to the central role of algorithmic personalisation - or ' recommending ' - by platforms in content dissemination in online environments and the problems to which this contributes. Focusing on recommender systems, i.e. the mechanism by which content is recommended by platforms, provides an alternative regulatory approach that avoids many of the pitfalls with addressing the hosting of content itself. This paper therefore explores motivations and considerations for regulating the use of recommender systems by online platforms. In doing so, this paper establishes a typology of online recommending, sets out various problems and consequences of recommending, and argues that recommending content is not one of the three activities for which information society service providers are afforded liability protections under the E-Commerce Directive. To address the identified problems and fill this legal gap, this paper proposes some principles for future regulation, and discusses approaches to oversight and compliance that could work with these principles.