Why your AI sounds confident when it's wrong

Models are trained to continue text that looks likely. Certainty is often a side effect of fluent style. When the stakes rise, ask for uncertainty, alternatives, and what would falsify the answer — then check outside the chat.