Prior research has established that various aspects of the user’s situation, collectively called the user model, affect what information is relevant. The purpose of our research is to refine this idea by exploring how different aspects of the user model are salient for different question types. Our methodology follows tradition in studying real intermediary elicitations for clues about what aspects of the user model are important, except that we analyze how this differs across question types. We find that there are more elicitations about the background of the user’s task and about the relevance of particular information for longer-answer questions than for short-answer questions, but surprisingly, no more elicitations regarding the sufficiency of particular information. The practical application of our research is to guide human or automated respondents to focus on the user details that are most important for different types of question.