We’ve heard story after story about people becoming obsessed with AI chatbots and experiencing profound breaks with reality.
But there was always an ambiguity: was AI use pushing people into psychosis, or were people already suffering from mental health problems seeking out AI as a means to cope?
Now, new research has provided an important clue. As PsyPost reports, scientists have found that people who use AI chatbots tend to experience higher levels of psychological distress compared to individuals who don’t. It doesn’t quite establish causation, but it’s a glaring sign of correlation — and the implications could be deep.
“Humans have an innate need to build and maintain meaningful relationships, and in today’s digital world, many of these connections increasingly unfold through technology,” author and Tampere University senior research fellow Iina Savolainen told PsyPost.
“With the rise of AI, more people are using social chatbots to explore new forms of communication or to seek companionship, emotional support, or simply, everyday interaction,” she continued. “Yet despite this growing trend, surprisingly few empirical studies have examined who uses these tools and how such usage relates to well-being.”
As detailed in a new paper published in the Journal of Social and Personal Relationships, Savolainen and her colleagues at Tampere analyzed data from a 2023 longitudinal study. The data included responses from 5,663 adults across six European countries — including Finland, France, Germany, Ireland, Italy, and Poland — who filled out an online survey assessing their use of “chatbot friends” services offered by companies like Replika and My AI.
The team measured the level of psychological distress using a 38-item measure of mental well-being called the Mental Health Inventory, and compared outcomes between those who reported using the AI chatbots and those who didn’t.
It was quickly clear that there was a correlation.
“The cross-cultural consistency was striking; social chatbot use was related to poorer mental well-being in all six countries,” Savolainen told PsyPost.
“Taken together, these results suggest that social chatbot use may emerge as a response to emotional or social challenges rather than as a tool that inherently improves well-being,” she added. “This does not mean chatbots can’t be helpful, but it reminds us that the dynamics of use are complex and may reflect underlying needs that technology alone cannot fully address.”
Importantly, the researchers admitted that the study only allowed them to “discuss associations” and not infer a cause-and-effect relationship. In other words, while it’s an intriguing link, we still don’t know if social chatbot usage actually causes negative mental health outcomes, or if the explanation is something more complex.
Another limitation of the study is the age of the dataset itself. The original survey was conducted in late 2023, less than a year after the launch of OpenAI’s ChatGPT, raising the possibility that trends may have shifted since then as more people have been drawn to the tech.
“Future research is needed to make more robust conclusions on social chatbot usage and their potential in supporting human health and well-being,” the paper reads. “Moreover, the possibility that these technologies are of particular interest to those who are already in a vulnerable position warrants investigation.”
As researchers continue to study the subject, they’re warning of the potential dangers of social chatbot usage, particularly among younger people.
In extreme cases, we’ve already found that an infatuation with AI models has played a role in a growing number of deaths, leading to lawsuits aimed at companies including Character.AI and OpenAI that are still playing out in court.
“We are living a unique time of human-computer interaction, as different technologies are becoming more sophisticated and culturally embedded,” Savolainen told PsyPost.
“As chatbots diversify and become more personalized, understanding the evolving nature of these relationships will be crucial,” she added.
More on AI chatbot use: A Startling Proportion of Teens Now Prefer Talking to AI Over a Real Person