Commentary driver training involves teaching drivers how to verbally acknowledge their perceptual and cognitive processes while driving, and has been shown to improve performance in driving-related tasks. However, those studies demonstrating benefits of commentary training have not done so under conditions of live commentary, which is the typical protocol used with advanced drivers. In the current study we present the results of 2 experiments that show that producing a commentary can actually slow responses to hazards on a concurrent hazard perception task. In Experiment 1, participants producing a live commentary showed significantly longer hazard response times than an untrained, silent, control group. In Experiment 2, a shorter, clipped commentary was introduced to attempt to reduce the demands placed upon participants. However, both the clipped and full commentary conditions showed reduced accuracy and longer response times, relative to a silent condition, and no difference was observed between the 2 types of commentary. Analysis of eye movements in both experiments revealed that fixation durations were shorter when a commentary was produced but time to first fixate the hazard was not affected. This suggests that commentaries encourage more active interrogation of the visual scene, but that this can be detrimental to performance in average drivers.