While exoskeleton technology is becoming more and more common for gait rehabilitation in children with neurological disorders, evaluation of gait performance still faces challenges and concerns. The reasoning behind evaluating the spinal locomotor output is that, while exoskeleton's guidance forces create the desired walking kinematics, they also affect sensorimotor interactions, which may lead to an abnormal spatiotemporal integration of activity in particular spinal segments and the risk of abnormalities in gait recovery. Therefore, traditional indicators based on kinematic or kinetic characteristics for optimizing exoskeleton controllers for gait rehabilitation may be supplemented by performance measures associated with the neural control mechanisms. The purpose of this study on a sample of children was to determine the basic features of lower limb muscle activity and to implement a method for assessing the neuromechanics of spinal locomotor output during exoskeleton-assisted gait. To this end, we assessed the effects of a robotic exoskeleton (ExoAtlet Bambini) on gait performance, by recording electromyographic activity of leg muscles and analyzing the corresponding spinal motor pool output. A slower walking setting (about 0.2 m/s) was chosen on the exoskeleton. The results showed that, even with slower walking, the level of muscle activation was roughly comparable during exoskeleton-assisted gait and normal walking. This suggests that, despite full assistance for leg movements, the child's locomotor controllers can interpret step-related afferent information promoting essential activity in leg muscles. This is most likely explained by the active nature of stepping in the exoskeleton (the child was not fully relaxed, experienced full foot loading and needed to maintain the upper trunk posture). In terms of the general muscle activity patterns, we identified notable variations for the proximal leg muscles, coactivation of the lumbar and sacral motor pools, and weak propulsion from the distal extensors at push-off. These changes led to the lack of characteristic lumbosacral oscillations of the center of motoneuron activity, normally associated with the pendulum mechanism of bipedal walking. This work shows promise as a useful technique for analyzing exoskeleton performance to help children develop their natural gait pattern and to guide system optimization in the future for inclusion into clinical care.