The growing evidence on psychotropic drug safety in pregnancy has been possible thanks to the increasing availability of real-world data, i.e. data not collected in conventional randomised controlled trials. Use of these data is a key to establish psychotropic drug effects on foetal, child, and maternal health. Despite the inherent limitations and pitfalls of observational data, these can still be informative after a critical appraisal of the collective body of evidence has been done. By valuing real-world safety data, and making these a larger part of the regulatory decision-making process, we move toward a modern pregnancy pharmacovigilance. The recent uptake of real-world safety data by health authorities has set the basis for an important paradigm shift, which is integrating such data into drug labelling. The recent safety assessment of sodium valproate in pregnant and childbearing women is probably one of the first examples of modern pregnancy pharmacovigilance.