Artificial intelligence companions are increasingly being used for emotional support, friendship and even romantic interaction. But a new study suggests that for some users - particularly those with limited social networks - these AI relationships may actually reinforce loneliness rather than relieve it.
Researchers from Stanford University examined how people use AI companions on Character.AI, a platform that allows users to create and interact with persona-based chatbots. Their findings, published in Nature Human Behaviour, indicate that the relationship between AI companionship and psychological well-being is far more nuanced than often assumed.
Human connection still matters
The study analysed survey responses from 1,131 Character.AI users in the United States, alongside nearly 465,000 messages from more than 4,600 donated chat sessions. Participants were asked why they used AI companions, how intensely they interacted with them and how willing they were to discuss sensitive personal issues, ranging from emotional distress to substance use and suicidal thoughts.
Although fewer than 12 percent of participants initially identified companionship as their primary reason for using the chatbots, more than half described their AI counterpart as a "friend", "companion" or "romantic partner". More than 80 percent of the donated conversations focused on emotional or social support.
The researchers found that intensive chatbot use was not inherently associated with poorer well-being. Instead, outcomes depended largely on users' motivations and circumstances. Participants who used AI companions primarily for companionship and who also had relatively small offline social networks reported significantly lower levels of psychological well-being. The same pattern was observed among users who frequently disclosed highly personal information to their AI companion.
Not adressing underlying causes
According to the researchers, AI companions may provide an immediate sense of connection without offering the reciprocal emotional exchange that characterises human relationships. Because these systems are designed to maximise engagement, they may encourage prolonged interaction without addressing the underlying causes of loneliness.
The research team is now investigating which aspects of human-AI interaction contribute most strongly to these effects. Potential safeguards under consideration include usage limits and mechanisms that encourage users to seek support from family members, friends or mental health professionals when conversations indicate emotional vulnerability.
Evidence remains mixed
The findings add to a growing body of research on AI companions, but they do not settle the debate. Earlier this year, researchers in South Korea reported that regular interactions with an AI-powered social chatbot reduced loneliness by an average of 15 percent and social anxiety by 18 percent among students who interacted with the system several times a week over a four-week period.
The researchers concluded that social chatbots may offer meaningful emotional support, particularly for people who struggle with face-to-face interactions, while emphasising that further research is needed to better understand their long-term benefits and limitations.
Growing role in mental health
At the same time, AI is playing an increasingly important role in how young people seek mental health support. Earlier research found that one in eight American adolescents and young adults used AI chatbots for emotional or psychological advice. A more recent study published in JAMA Pediatrics suggests that figure has already risen to 19.1 percent, with more than 40 percent of users consulting chatbots at least monthly.
Nearly two-thirds of respondents said they had not told parents, friends or others about their chatbot use, even though almost 92 percent considered the advice helpful. Researchers caution, however, that perceived helpfulness should not be confused with clinical effectiveness. Together, these findings highlight both the growing role of AI in the mental health ecosystem and the urgent need for safeguards, transparency and continued access to professional care as conversational AI becomes an increasingly common source of emotional support.
