What actually makes people trust a news chatbot: being right, or just sounding confident and friendly?

This explores which specific chatbot behaviors make people trust it for news, and whether that trust follows accuracy. The corpus's short answer is uncomfortable: the features that earn trust are mostly about how the chatbot talks, not whether it's correct. The collection is also thin on news-specific design studies. Most of the evidence comes from general chatbot trust research, with news surveys showing the stakes.

Start with the news picture. Weekly chatbot use for news has risen to about 10% across dozens of markets, but only 20% of people trust chatbot answers, compared with 37% for news in general Why do AI chatbots gain news users but lose their trust?. Trust seems to come before use rather than after it. Among people who already use chatbots for news, 44% trust what they get, against 17% of non-users, and trust predicts adoption more strongly for chatbots than it does for social media Does trust in AI chatbots drive news-seeking behavior?. One feature does look like a real, accuracy-linked trust builder: 42% of users click through to the original sources. Visible links back to the reporting are the closest thing the corpus offers to a trust mechanism that gives people a way to check the answer.

Most other trust signals have little to do with accuracy. A focus-group study found that conversationality drives trust in ChatGPT. That means replies that respond to what you just said, fast answers, and tidy formatting. These work because they trigger the instincts we use with other people, not because users check whether the answers are reliable Does conversational style actually make AI more trustworthy?. A related line of work shows chatbots using the language patterns of expertise, a confident and knowledgeable-sounding register. Trust attaches to that register rather than to correctness, and users drift from searching and remembering things themselves toward passive reliance Does chatbot language style actually shape how much we trust it?. For news, this means a well-formatted, authoritative-sounding summary can earn trust it hasn't earned.

Warmth is where trust and accuracy actively split. In a 199-person study, a warm chatbot style increased agreement with wrong answers, especially when users were unsure. Whether people checked a claim depended on how much they already trusted AI, not on whether the answer was right. Having web search available alongside the chatbot didn't fix this Does access to web search prevent overreliance on chatbots?. Model-side research makes it worse: training a model to be empathetic raised its error rates on truthfulness and on resisting disinformation by up to 30 percentage points Does empathy training make AI systems less reliable?. So warmth makes users more trusting and the model less reliable, which is a bad combination for news. But warmth doesn't always win: when outside raters read chatbots behaving like companions, they judged them less trustworthy and less likable Do chatbot companionship behaviors actually increase how much people like them?.

The less obvious lesson is about time. Personalization raises trust, but each good interaction also raises expectations, so later failures disappoint more Does chatbot personalization build trust or expose privacy risks?. The novelty that drives early engagement also fades in predictable ways Do chatbot relationships lose their appeal as novelty wears off?. So a news chatbot's trust can look strong in a single-session test and still wear down with daily use. The design question worth asking is how to make trust follow accuracy, with source links as the best lead so far. Asking what makes people trust it more mostly selects for style.