A Hacker News discussion reveals growing frustration with people who blindly trust LLM responses as objective truth. Users describe colleagues citing AI outputs like they're autonomous beings, even when presented with contradictory evidence. The problem extends beyond casual conversation - people are using hallucinated legal cases in lawsuits and AI-generated content in professional settings.

Commenters suggest various approaches to address this trust issue. Some recommend demonstrating AI's sycophantic nature by showing how it flip-flops on topics. Others emphasize treating AI like any other source - holding users accountable for verification regardless of origin. One commenter compares AI trust to historical bad advice acceptance, arguing that responsibility ultimately falls on the user.

Several users note this reflects a broader societal trust crisis, not just an AI problem. The discussion highlights how AI's convincing delivery makes misinformation harder to detect. As one commenter observes, people need to experience being 'hallucinated at' to understand the limitations. The consensus suggests gentle reframing - consistently using 'it' instead of 'he/she' for AI, encouraging gut checks, and maintaining the same standards for AI-generated content as any other source.