In Kennedy v. Louisiana, the Supreme Court noted that the problem of “unreliable, induced, and even imagined child testimony” creates a “special risk of wrongful execution in some child rape cases.”‘ Indeed, empirical research has repeatedly demonstratedproblems with accuracy in children’s accounts of their own experiences. Although the research and commentary inthis area hasfocused on how allegations of child sexual abuse are addressedinthe criminaljustice system, these studies have much broaderimplications. every year, stateofficials conductmillions of interviews with children in the context ofchild welfare investigations. These investigations have serious consequences for families-for instance, they can lead to the placement of a child in foster care or the termination of parental rights. This article examines the reliability of child welfare determinationsby looking at a subset of the information investigators consider: children’s statements about theirpast experiences. First, this article reviews empirical research on suggestibility, lie-telling behavior, and the capacity of adults to detect lies in children. The article then examines the impact of structural features in the child welfare system, and posits that these structural features do not facilitate the proper evaluation of child statements. The article concludes by proposing legal reforms to improve the reliability of child welfare determinations. Ultimately, this article aims to defend the proposition that caring deeply about children and their safety does not necessarily mean the child welfare system should rely on what children say.