A new approach to interacting with artificial intelligence (AI) systems involves prompting them to identify their own uncertainties. This technique helps users understand the limitations of AI responses and recognize when more information is needed. By asking the AI, "What are you unsure about in your answer?" users can uncover assumptions the AI has made and areas where its knowledge might be incomplete. For example, when an AI was asked to plan a three-day family trip, it created an itinerary that seemed well-structured. However, it didn’t consider specific details about the family’s preferences or potential disruptions like weather or traffic. When asked to identify uncertainties, the AI acknowledged concerns about the itinerary’s pacing and the lack of information about the family’s needs and travel arrangements. In another instance, the AI was asked whether to repair or replace an older laptop. It provided a general framework for making the decision but didn’t have specific details about the laptop’s condition or repair costs. When prompted to identify uncertainties, the AI admitted it lacked information about the laptop’s issues and repair costs, which are key factors in making an informed decision. This method can be expanded by asking the AI to highlight the most likely uncertainty that could change its recommendation or what specific information is needed for a better response. This is especially helpful in decision-making scenarios, where users can first receive an initial recommendation and then explore the weakest parts of the response. However, it's important to remember that the AI itself generates its description of confidence, and it may not always accurately identify its own limitations. For decisions involving critical facts, this technique should not replace checking reliable sources.