The mobile psychomotor vigilance task (PVT) has been found to be a valid predictor of cognitive fatigue. However, absolute reaction time (RT) recorded by mobile PVT is inaccurate. This is concerning as participant RTs are used in the analysis of PVT results. This paper aims to characterise this problem and assess the margin of error across common iOS software frameworks. A novel Arduino test instrument was developed to simulate a user’s reaction, providing a ground truth for the RT. We found in our experiments that there is between a 29.57% and 48.58% increase over the ground truth RT in the iOS implementations tested. These are significant overestimations that will affect the validity of the outcome metrics for any mobile PVT study participants.