Artificial intelligence (AI) can support mammography reading in several ways, including as a second reader and via decision support. In this study, 758 mammograms were read by human radiologists with no AI support and read again 6 weeks later, this time with the AI-predicted diagnosis of positive or negative for cancer. AI support resulted in a decrease in sensitivity and increase in specificity compared to human-only reading. The prevalence of cancer in the sample was higher than in the general population (1:1 cancer to healthy ratio), which also affected the readers' accuracy.