As artificial intelligence (AI)-powered chatbots become more popular and more humanlike in the way they communicate, users are increasingly turning to them as trusted sources of information on topics ranging from personal finance to health. That may be problematic, according to Penn State researchers, who found in a recent study that the more conversational an AI chatbot is, the more likely users are to trust it—even when it shares misinformation. The team said verification options may help rebuild needed skepticism.

AI-generated texts are increasingly perceived as human, but people can still recognize human writing as such. However, this distinction becomes much more difficult once AI is explicitly instructed to come across as human. This is shown by research involving, among others, Tilburg University, published in the journal Nature Communications. Several experiments involving more than 3,000 participants examined how human and empathetic people perceive relationship advice generated by AI and humans.