Although stroke is one of the world’s leading causes of death and disability, and more than a thousand candidate neuroprotective drugs have been proposed based on extensive in vitro and animal-based research, an effective neuroprotective/restorative therapy for ischaemic stroke patients is still missing. In particular, the high attrition rate of neuroprotective compounds in clinical studies should make us question the ability of in vitro models currently used for ischaemic stroke research to recapitulate human ischaemic responses with sufficient fidelity. The ischaemic stroke field would greatly benefit from the implementation of more complex in vitro models with improved physiological relevance, next to traditional in vitro and in vivo models in preclinical studies, to more accurately predict clinical outcomes. In this review, we discuss current in vitro models used in ischaemic stroke research and describe the main factors determining the predictive value of in vitro models for modelling human ischaemic stroke. In light of this, human-based 3D models consisting of multiple cell types, either with or without the use of microfluidics technology, may better recapitulate human ischaemic responses and possess the potential to bridge the translational gap between animal-based in vitro and in vivo models, and human patients in clinical trials.