This study examined the effect of safety behavior on belief change during a behavioral experiment. Highly spider fearful participants (N = 126) evaluated a targeted negative belief about spiders during a brief behavioral experiment with a live tarantula. Participants were randomly assigned either to use or not use safety gear during the session. Results demonstrated that after the behavioral experiment, targeted negative beliefs were significantly lower in the safety gear condition than in the control condition. Both conditions benefited from comparable improvements across a broader constellation of negative spider-related beliefs. Safety gear facilitated closer approach to the spider during the session; however, participants who did not use safety gear experienced greater improvement in perceived control. These findings suggest that safety behavior need not impair corrective learning during cognitive-behavioral interventions and that it might indeed enhance it. Results are discussed in terms of cognitive-behavioral treatments for anxiety disorders and of the role of safety behavior therein.