High-throughput screening of libraries containing compounds of 'drug-like' molecular weight has frequently resulted in no or poor drug candidates, especially when screening against demanding drug targets such as protein-protein interactions. Fragment-based lead discovery and optimization has evolved as a promising solution to this problem by combining the universal adaptability of low-molecular-weight fragments with immediate structural information on fragment binding modes. This review focuses on nuclear magnetic resonance (NMR) fragment screening techniques, which provide a unique combination of medium-throughput, direct binding site information and broad applicability. The utility and exemplary data of chemical shift-detected NMR fragment screening applied to the challenging protein-protein interaction target PDZ domains are summarized.