Recommendation agents are deployed to give online consumers advice on products. This study focuses on how two demographic interface characteristics of online recommendation agents – ethnicity and gender – influence the way consumers agree with the product recommendations offered by anthropomorphic (humanoid) recommendation agents. Because consumers may not always straightforwardly self-report their true perceptions about entities that differ from them in their ethnicity and gender, this study applies neuroimaging methods (fMRI) to understand how the design of online recommendation agents can include anthropomorphic interfaces with different ethnicity and gender to enhance the interaction between consumers and agents. Subjects who either fully matched or fully mismatched with the ethnicity and gender of recommendation agents were asked to indicate their agreement with the advice provided by the recommendation agents while their brain activities were observed in an fMRI scanner. The results show that there is only activation in brain areas of intense emotion (amygdala) and fear of loss (insular cortex) when subjects disagree with a recommendation agent that does not match their ethnicity and gender, while there is no activation for recommendation agents that match their ethnicity and gender. The fMRI results suggest that ethnicity and gender mismatch spawns strong emotional responses in the brain, particularly among women.