In this paper, a semantic approach to building multi-platform 3D content is proposed. The presented solution is intended to enable flexible and efficient creation of 3D presentations covering a wide range of target platforms – visualisation tools, content representation languages and programming libraries. Referring to the semantics of particular content elements can facilitate conceptual knowledge-based content creation at arbitrarily high levels of abstraction, and it can improve indexing, searching and analysis of 3D content in a variety of application domains on the web.