Is using a robot pal to talk about your worries or ask for advice a good idea?

A new study from researchers in Denmark suggests those questions may have troubling answers, though they are far from definitive.

A chatbot is a software application that simulates human conversation. You might text and ask it to suggest a hairstyle. Or prompt it to write a poem. Some people use them as companions.

With more than 1.3 billion downloads since 2022, ChatGPT is now widely used.

The Danish researchers wanted to test what had only been reported anecdotally. They searched more than 10.7 million psychiatric patient records for clinical notes that mentioned “ChatGPT,” “chatbot” or had a related spelling.

They found 38 cases in which chatbot use appeared to have potentially added to mental harm, such as feeding delusions, suicidal thoughts, eating disorders and obsessive-compulsive disorder.

In 11 cases, chatbots reinforced a patient’s false beliefs. Six patients showed signs of self-harm tied to chatbot interactions. For five patients with eating disorders, chatbots enabled calorie-counting or food restriction.

While those aren’t large numbers, researchers say the true numbers are likely much higher. Most clinicians do not consistently ask patients about chatbot use. And if they do, it might not end up in a patient’s chart. The researchers asked only about ChatGPT, not its many competitors.

For now, the researchers suggest that mental health professionals talk to patients about their chatbot use, because some would benefit from either reducing their reliance on them or skipping them entirely.