The failure of AI （artificial intelligence）assistant services is essentially an algorithmic failure, and the "black box" nature of the algorithmic decision-making process exacerbates consumers' negative reactions to service failures. A between-subjects experimental approach was adopted to explore the impact mechanism of enhancing algorithmic explainability from the perspective of perceived control through explainable artificial intelligence methods (such as post hoc explanations) on consumer behavior in the context of service failure, as well as the boundary conditions. It is found that when algorithmic explainability is enhanced (compared to the control), consumers' continued intention to use despite service failure is improved, and consumers' perceived control plays a partial mediating role in this process. However, the above effects are not significant when the anthropomorphism level of the AI assistant is low (compared to high).