Landmarks support the structuring of environmental information into cognitive conceptual units, they have the potential to identify uniquely pertinent intersections for route following, and they disambiguate spatial situations at complex intersections. Not using them in automatically generated route directions is a violation of cognitive ergonomics. While we have made great progress on the one hand in characterizing and on the other hand in mining potential landmarks, viable data structures that incorporate their cognitive conceptual functions in route directions are poorly developed. The present article closes this gap by providing a representation based on the OpenLS standard that allows for capturing the semantics of landmarks. In this data structure, the cognitive conceptual essence of a landmark is represented allowing for generating route directions automatically and imbuing street network data with cognitively meaningful elements.