General practitioners (GPs) do not provide enough preventive care. Nonetheless, without a detailed understanding of the logical processes that underlie their practices, it remains difficult to develop effective means of improvement. Their relationship to knowledge is one of three elements that strongly structure GPs’ preventive work (together with the doctor–patient relationship and the organisation of their professional space).The objective of this article was to explore the question of GPs’ relationship to knowledge about prevention. In 2010–2011, semi-directive interviews with a diverse sample of 100 GPs practising in the Paris metropolitan area were conducted. These interviews were coded according a reading grid that was developed collectively and analysed in the framework of grounded theory. The cognitive universe of GPs is neither homogeneous nor stable. It is composed of biomedical knowledge (delivered via guidelines, the professional press, opinion leaders and pharmaceutical companies), clinical knowledge (fed by individual situations from their daily experience and often conflicting with epidemiologic reasoning and data) and lay knowledge (from folk culture). Plunged into this complex cognitive universe that is difficult for them to master, doctors construct their own idiosyncratic preventive style by themselves, mostly in isolation. Two types of actions emerged as likely to help GPs better appropriate preventive knowledge: clarification of scientific data (especially from epidemiology and the social sciences) but also development of a collective analysis of the cognitive work required to integrate the different types of knowledge mobilised daily in their preventive practices.