Robot-mediated therapy can help improve walking ability in patients following injuries to the central nervous system. However, the efficacy of this treatment varies between patients, and evidence for the mechanisms underlying functional improvements in humans is poor, particularly in terms of neural changes in the spinal cord. Here, we review the recent literature on spinal plasticity induced by robotic-based training in humans and propose recommendations for the measurement of spinal plasticity using robotic devices. Evidence for spinal plasticity in humans following robotic training is limited to the lower limbs. Body weight-supported (BWS) robotic-assisted step training of patients with spinal cord injury (SCI) or stroke patients has been shown to lead to changes in the amplitude and phase modulation of spinal reflex pathways elicited by electrical stimulation or joint rotations. Of particular importance is the finding that, among other changes to the spinal reflex circuitries, BWS robotic-assisted step training in SCI patients resulted in the re-emergence of a physiological phase modulation of the soleus H-reflex during walking. Stretch reflexes elicited by joint rotations constitute a tool of interest to probe spinal circuitry since the technology necessary to produce these perturbations could be integrated as a natural part of robotic devices. Presently, ad-hoc devices with an actuator capable of producing perturbations powerful enough to elicit the reflex are available but are not part of robotic devices used for training purposes. A further development of robotic devices that include the technology to elicit stretch reflexes would allow for the spinal circuitry to be routinely tested as a part of the training and evaluation protocols.