Vast swaths of geographic information are produced by non-professional contributors using online collaborative tools. To extract value from the data, creators and consumers alike need some degree of consensus about what the entities of their domain of interest are and how they are related. Traditional information communities, such as government agencies, universities, and corporations, have devised informal and formal mechanisms to reduce the misinterpretation of the data they rely on, curating vocabularies, standards, and, more recently, formal ontologies. Because of the decentralized, fragmented nature of peer production, semantic agreements are more difficult to establish and to document in volunteered geographic information (VGI), severely limiting the re-usability and, ultimately, the value of the data. This paper provides an overview of the semantic issues experienced in VGI, and what potential solutions are emerging from research in geo-semantics and in the Semantic Web. The paradigm of Linked Data is discussed as a promising route to handle the semantic fragmentation of VGI, reducing the friction between data producers and consumers.