In this study, we investigated how children interact with voice assistants, particularly focusing on what kinds of questions they ask and how they react to the responses obtained. We recorded 3-to 10-year-old children's (N = 43) spontaneous interactions with Amazon Alexa, and analyzed the questions they asked, as well as how they adjusted their information search based on the responses received. Our results confirm previous work in showing that children's questions are mostly information-seeking, yet the type of questions children ask also depends on their age and familiarity with voice assistants. For example, children who are younger and less familiar with voice assistants are more likely to ask questions about themselves and their environment (e.g., "What is my sister's name?"). We also show for the first time that, even though all children are sensitive to the relevance and accuracy of voice assistants' responses to a certain extent, older children are more likely to change the topic and type of the questions asked upon receiving irrelevant or uninformative responses.This study shows that, with age and familiarity, children become more sensitive to the behavior, informativeness, and constraints of artificial agents, growing into adaptive and sophisticated technology users.