The current studies aimed to measure and modify approach biases for physically active lifestyle cues. In Study 1, 136 participants who, on average, met physical activity guidelines completed an implicit association task (IAT) to measure approach bias for words representing a physically active or inactive lifestyle. Approach bias scores were positive overall, indicative of an implicit motivation to approach physically active lifestyle cues. Study 2 explored whether pre-existing approach biases could be increased using approach training (pairing active words mostly with approach words) or decreased using avoid training (pairing active words mostly with avoidance words). The sample consisted of 73 participants who self-identified as being insufficiently active and, on average, reported levels of physical activity that were below recommended guidelines. Approach bias for physically active lifestyle cues was stronger for the approach group than the avoid group at post-training. However, training effects were moderated by the strength of the pre-existing approach bias; the group difference in approach bias was more apparent among participants with either a moderate or a strong pre-existing approach bias. The results are consistent with dual-process models which highlight the importance of non-conscious processes in health behaviour and intervention, and suggest that cognitive bias modification may be a useful pathway for promoting physical activity engagement.