Background The Lokomat is a robotic exoskeleton that can be used to train gait function in hemiparetic stroke. To purposefully employ the Lokomat for training, it is important to understand (1) how Lokomat guided walking affects muscle activity following stroke and how these effects differ between patients and healthy walkers, (2) how abnormalities in the muscle activity of patients are modulated through Lokomat guided gait, and (3) how temporal step characteristics of patients were modulated during Lokomat guided walking. Methods Ten hemiparetic stroke patients (>3 months post-stroke) and ten healthy age-matched controls walked on the treadmill and in the Lokomat (guidance force 50%, no bodyweight support) at matched speeds (0.56 m/s). Electromyography was used to record the activity of Gluteus Medius, Biceps Femoris, Vastus Lateralis, Medial Gastrocnemius and Tibialis Anterior, bilaterally in patients and of the dominant leg in healthy walkers. Pressure sensors placed in the footwear were used to determine relative durations of the first double support and the single support phases. Results Overall, Lokomat guided walking was associated with a general lowering of muscle activity compared to treadmill walking, in patients as well as healthy walkers. The nature of these effects differed between groups for specific muscles, in that reductions in patients were larger if muscles were overly active during treadmill walking (unaffected Biceps Femoris and Gluteus Medius, affected Biceps Femoris and Vastus Lateralis), and smaller if activity was already abnormally low (affected Medial Gastrocnemius). Also, Lokomat guided walking was associated with a decrease in asymmetry in the relative duration of the single support phase. Conclusions In stroke patients, Lokomat guided walking results in a general reduction of muscle activity, that affects epochs of overactivity and epochs of reduced activity in a similar fashion. These findings should be taken into account when considering the clinical potential of the Lokomat training environment in stroke, and may inform further developments in the design of robotic gait trainers. Electronic supplementary material The online version of this article (doi:10.1186/s12984-017-0244-z) contains supplementary material, which is available to authorized users.