Fatigue is an important design criterion for welded structures subjected to cyclic loading. Several approaches for fatigue strength assessment have been developed which are either based on Woehler S–N curves and damage accumulation rule or on crack propagation law. The paper briefly reviews the different approaches, highlighting their advantages and limitations. In this connection, the problematic distinction between crack initiation and propagation phases is discussed, followed by considerations about some parameters which have large influence on the fatigue behaviour of welded joints but are considered differently in the approaches, such as plate thickness and stress gradient effects, multiaxial stress states, welding-induced distortions and residual stresses. Finally, ways of improving the fatigue behaviour of welded structures, either during design by reducing the stress concentration or during fabrication by improved quality or post-weld treatment or else by special material characteristics, are addressed. Emphasis will be placed on recent developments and challenges for the future from a personal perspective of the author.