The appearance anthropomorphism of service robots is mainly aimed at gaining recognition and trust from consumers, but as appearance anthropomorphism reaches a certain critical point, it will bring negative results. In the context where the critical point has not yet been clear, this paper explores the interaction effect of robot anthropomorphism and task types on consumer preferences based on the anthropomorphism theory. The study conducted three experiments to demonstrate that task types can trigger consumers’ varying expectations for the behavior anthropomorphism of service robots. High interaction tasks lead consumers to have high expectations for behavior anthropomorphism, and at this time, consumers prefer robots with high (vs. low) appearance anthropomorphism, with trustworthiness playing a mediating role; on the contrary, low interaction tasks lead consumers to have low expectations for behavior anthropomorphism, and at this time, consumers have a feeling of eeriness towards robots with high (vs. low) appearance anthropomorphism, being more inclined to accept robots with low (vs. high) appearance anthropomorphism. This paper not only enriches the research context of the anthropomorphism of service robots and deepens the anthropomorphism theory, but also provides important insights for enterprises to effectively apply service robots with anthropomorphism.