Attentional bias towards negative faces means that, compared to positive or neutral faces, negative faces could guide attention to themselves more efficiently. Attentional bias helps people adapt more successfully in response to danger, and also contributes causally to anxiety or depression vulnerability. Therefore, there is necessity to find out an efficient way to affect attentional bias. However, most previous study about attentional bias examined whether attentional bias depend on perceptual or emotional differences, while little is known about the mechanism of how to affect attentional bias using top-down strategies. Previous studies have found that automatic emotion regulation can affect attention allocation, cognition, emotional experience, and behavior during process through which emotions are generated, without making much effort. Studies also showed that automatic emotion regulation affect emotion by a top-down way. The present study aims to examine whether automatic emotion regulation, a top-down and effortless strategy, could influence attentional bias towards negative faces, by adopting emotional flanker task and the idiom matching task. A 2 (priming type: neutral, emotion regulation) × 2 (target face emotion type: positive, negative) × 2 (congruency: congruent, incongruent) between-subject experiment design was carried out. There were 10 blocks in total and each block consisted of 2 successive trials of the matching task and 16 successive trials of emotional flanker task. Five successive blocks used idioms concerned with emotion regulation and the other half of the blocks used neutral idioms. The sequence of priming types was counterbalanced between subjects. In the idiom matching task, three Chinese four-character idioms were presented until participants made their choice between the lower two idioms to match the meaning of the upward idiom. In the emotional flanker task, a fixation cross was presented for 500 ms, immediately followed by a target face with 2 flanking faces beside it. Participants were instructed to indicate which kind of emotion the target face expressed. The faces would disappear until participants made the choice or after 1500 ms. Then the screen went blank for an inter-trial interval of 1000 ms. The result showed that in neutral priming condition, participants responded faster to congruent trials than incongruent trials with positive target; however, when the target was negative, the RTs in congruent trials than incongruent trials showed no difference. In emotion regulation priming condition, for both positive and negative targets, RTs in congruent trials and incongruent trials showed no difference. These results suggested that automatic emotion regulation, primed by the idiom matching task, could affect attentional bias toward negative faces, and extended previous studies suggesting that faces expressing negative emotion can capture attention more efficiently.