Have you ever wondered how much you trust what you read online? With the rise of large language models (LLMs) generating everything from news headlines to poems, this question of trust becomes even more critical. New research dives deep into how we perceive the trustworthiness of LLM-generated content, revealing some surprising results. A study from MIT Lincoln Laboratory explored how explanations accompanying AI-generated news headlines influenced people’s trust. Participants were asked to rate headlines, both with and without justifications for why the AI chose them. The findings suggest that people generally tend to trust AI responses, especially when presented in isolation. However, when given the chance to compare different headlines side-by-side, a clear preference emerged. Headlines with explanations, particularly those offering reasoning *after* the headline, were deemed more trustworthy. Interestingly, intentionally false explanations lowered trust when compared to other options. This highlights the importance of transparency—even if the explanation isn’t perfect, it helps build trust. This research has important implications for the future of AI and human interaction. As LLMs become more integrated into our daily lives, understanding the nuances of trust will be key to effective collaboration. While explanations boost trustworthiness, this study underscores how easily we can be misled by isolated, deceptively convincing responses. Further research is needed to create better methods of identifying such deceptive responses and develop systems that promote both trust and accuracy in AI communication.