One of the aspects that policymakers should consider when promoting cycling is the route choice behaviour of current cyclists. This study develops a behaviourally realistic route choice model for different types of everyday cyclists and cycling trips. The analysis is based on a large-scale crowdsourced dataset of GPS trajectories including 134,169 trips from 6,523 cyclists. The model is estimated as a joint path-size logit model and accounts for a wide range of bicycle network attributes, such as bicycle infrastructure type, land use, surface type or cycle superhighways. The findings of the model reveal, for example, that infrequent cyclists feel less safe on large roads, but this effect can be accommodated with protected bicycle tracks. Interaction with other motorised and non-motorised transport modes is found to be a deterring factor for cyclists and they prefer scenic water and green areas over high-rise urban environments, especially on long trips. The model performs very well on a hold-out sample, also when considering the similarity between the observed and predicted route, not only their binary consistency. Finally, we formulate several policy measures relevant to promote cycling. Building long, continuous stretches of dedicated, protected bicycle infrastructure outside of the high-rise urban environments has the greatest potential to make cycling attractive.