The origin of fluorine is still a debated question. AGB stars synthesise this element and likely contribute significantly to its synthesis in the present-day Universe. However, it is not clear whether other sources contribute, especially in the early Universe. We discuss variations of the surface abundances of fluorine coming from our massive star models and compare them with available present-day observations. We compute the contribution of massive stars in producing 19F over metallicities covering the whole cosmic history. We used models in the mass range of 9Msol < Mini < 300Msol at metallicities from Pop III up to super-solar while accounting for the required nuclear network to follow the evolution of 19F during the core H- and He-burning phases. Results from models with and without rotational mixing are presented. We find that rotating models predict a slight depletion of fluorine at their surface at the end of the MS phase. In more advanced evolutionary phases, only models with an initial mass larger than 25Msol at metallicities Z > 0.014 show phases where the abundance of fluorine is enhanced. This occurs when the star is a WR star of the WC type. WC stars can show surface abundances of fluorine ten times larger than their initial abundance. However, we obtained that the winds of massive stars at metallicities larger than Z=0.006 do not significantly contribute to fluorine production, confirming previous findings. In contrast, very metal-poor rapidly rotating massive star models may be important sources of fluorine through the mass expelled at the time of their SN explosion. Observations of WC stars at solar or super-solar metallicities may provide very interesting indications on the nuclear pathways that lead to fluorine production in massive stars. The possibility of observing fluorine-rich CEMPs is also a way to put constrains in present models at very low metallicities.