The article draws on the first findings of the INFOCORE project to better understand the ways in which different types of media matter to the emergence, escalation or conversely, the pacification and prevention of violence. It makes the case for combining an interactionalist approach of media influence, which is centred on the effects of evidential claims, frames and agendas made by various actors over time, with greater sensitivity for the factors that make conflict cases so different. We argued that the specific role played by the media depends, chiefly, a) on the ways in which it transforms conflict actors' claims, interpretations and prescriptions into media content and b) their ability to amplify these contents and endow them with reach, visibility and consonance. We found significant variation in media roles across six conflict cases and suggest that they are best explained by four interlocking conditioning factors: (i) the degree to which the media landscape is diverse and free, or conversely, controlled and instrumentalised by conflict parties; (ii) societal attitudes to and uses of different media by audiences; (iii) different degrees of conflict intensity and dynamics between the conflict parties; (iv) the degree and nature of the involvement of regional and international actors. We argue that de-escalatory media influence will be most effective over the longer term, in settings of low intensity conflict and when tailored carefully to local conditions.