Surface nuclear magnetic resonance is a geophysical technique providing the ability to produce images of the subsurface water content profile. To produce reliable images, the physics of the excitation process must be accurately captured. Generally, the excitation is assumed to occur through a process known as on-resonance excitation. This assumes that the precession frequency of the magnetization is described by a single frequency at all locations in the subsurface and that the frequency of the applied magnetic field is set to this single frequency. However, several conditions can occur in which these assumptions may be violated. We explored two scenarios in which this is the case: Magnetic susceptibility contrasts lead to a distribution of precessional frequencies and (1) the transmit frequency is equal to the center of this distribution and (2) the transmit frequency is not equal to the center frequency of the distribution of precessional frequencies. Both of these scenarios leads to a condition known as off-resonance excitation. In this case, the spatial distribution of an excited signal in the subsurface may vary, and the measured signal’s amplitude and phase would be impacted. We have explored how off-resonance excitation, when incorrectly modeled as on-resonance excitation, impacts the predicted water content profiles for synthetic and field studies. We observed that neglected off-resonance effects leads to biased water content estimates and degrade the performance of the survey. Long pulse durations were observed to be sensitive to these effects, producing poor representations of the true subsurface water content profiles when off-resonance excitation was neglected, whereas short pulse durations were observed to alleviate these effects and produced accurate water content profiles.