Although technically challenging, detecting Earth-like planets around very low mass stars is in principle accessible to the existing velocimeters of highest radial-velocity (RV) precision. However, low-mass stars being active, they often feature dark spots and magnetic regions at their surfaces generating a noise level in RV curves (called activity jitter) that can severely limit our practical ability at detecting Earth-like planets. Whereas the impact of dark spots on RV data has been extensively studied in the literature, that of magnetic features only received little attention up to now. In this paper, we aim at quantifying the impact of magnetic fields (and the Zeeman broadening they induce) on line profiles, line bisectors and RV data. With a simple model, we quantitatively study the RV signals and bisector distortions that small magnetic regions or global magnetic dipoles can generate, especially at infrared wavelengths where the Zeeman broadening is much larger than that in the visible. We report in particular that the impact of magnetic features on line bisectors can be different from that of cool spots when the rotational broadening is comparable to or larger than the Zeeman broadening; more specifically, we find in this case that the top and bottom sections of the bisectors are anticorrelated, i.e. the opposite behaviour of what is observed for cool spots. We finally suggest new options to show and ultimately filter the impact of the magnetic activity on RV curves.