Stroke is the leading cause of complex disability. Indeed, 40% of person suffering from stroke present loss for motor function in their upper limp. Promising results have been shown with brain-computer interfaces (BCI) for improving motor rehabilitation. Recent meta-analysis have shown positive effects of BCI on motor rehabilitation of stroke patients. Most studies give a sensory feedback that can be either kinetic (robotic orthosis or neuromuscular electrical stimulation) or visual. The training seems to be more efficient when it includes somatosensory feedback. However, some issues need to be addressed to popularize BCI in hospitals. Two aspects in particular need to be improved: the personalisation of the training by taking into account the specificity of users (demographics data, cognitive and personality profiles, mental states) and the improvement of the classification performances including data of the profile and states of the user. Among the factors related to users, the somatosensory impairments related to the stroke certainly plays an important role in the motor recovery capacities, but this factor is not often taken into account. Considering user’s characteristics to improve the effects of BCI is a challenge that could be addressed with artificial intelligence (AI) algorithms. However this approach requires a better understanding of the factors affecting BCI-based post-stroke motor recovery in order to calibrate the algorithms used. In this presentation, we elaborate on the state-of-the-art of the individual factors that influence BCI motor rehabilitation. Also, we present the protocol of our study, which objective is to collect data regarding patients’ characteristics (demographic, cognitive, and personality) that would allow us to identify which patients best benefit from BCI-based motor recovery. Particular attention will be paid to the effect of stroke-related somatosensory impairment, fatigue and attention on post-stroke motor rehabilitation by BCI.