Humanitarian aid has rapidly emerged as a core component of modern peacebuilding and post-conflict reconstruction. However, some practitioners and policymakers claim that humanitarian assistance may actually prolong conflict. The current debate about the effect of humanitarian aid on conflict underspecifies causal mechanisms and takes place largely through case studies. I use a bargaining framework to argue that aid can inadvertently increase each combatant's uncertainty about the other side's relative strength, thereby prolonging civil war. I test my argument using panel data on cross-national humanitarian aid expenditures. From 1989 to 2008, increased levels of humanitarian assistance lengthen civil wars, particularly those involving rebels on the outskirts of a state. This result suggests that policymakers need to carefully consider whether the specific benefits provided by humanitarian aid outweigh the risk of prolonging civil conflicts, and to look for methods of disbursement that reduce that risk.