Recent work suggests that dishonesty results from ethical blind spots: people’s lack of attention to ethical information. In two experiments (one pre-registered) we used eye tracking to investigate when ethical blind spots emerge, and whether they can be reduced through a simple, non-invasive intervention. Participants reported a Target Digit indicated by a jittery cue that was slightly biased in the direction of another digit (the Second-Cued Digit), which could be either higher or lower than the Target Digit. Participants were paid more for reporting higher digits, and were not penalized for making mistakes, thus providing an incentive to cheat. Results showed that participants frequently made self-serving (and rarely self-hurting) mistakes by reporting the Second-Cued Digit when it was more valuable than the target. Importantly, they rapidly gazed at the digit that they would later report, regardless of whether this report was correct or a self-serving mistake. Finally, we were able to reduce or increase the number of self-serving mistakes by respectively increasing or reducing the visual saliency of the Target Digit. We suggest that increasing the visual saliency of morally desirable options is a promising cost-effective tool to curb dishonesty.