Original studyFrontiers in Psychology, 2026Read the paper →
AI companion apps such as Xingye and Maoxiang, which offer everyday chat and emotional companionship around the clock, have become part of daily life for many young people in China. Since 15 July 2026, national rules have required these services to watch for signs of emotional dependence and to remind users after two hours of continuous use. A new study suggests that loneliness tracks not only how much young adults use these apps, but whether they feel the apps have taken the place of real people.
A researcher at Yunnan Normal University surveyed 843 Chinese adults aged 18 to 29, all of whom had used a companion AI app in the previous three months. The study, published in Frontiers in Psychology, treated chatting with AI as a form of media use. The main analysis covered 783 people with complete data.
Participants were recruited through the online survey platform Wenjuanxing and received a small cash payment. Just over half were women, and most held a bachelor’s degree. The analysis allowed for age, education, living arrangements, how often people met friends or family in person, daily time spent with AI, and how long people had been using these apps.
Loneliness was measured with a short Chinese version of the UCLA Loneliness Scale, a widely used questionnaire. The study also measured what it calls self-attributed offline social displacement: a person’s own sense that time once spent face to face with friends or family now goes to chatting with AI. A third measure, relational interdependence, captured how far people define themselves through their close relationships.
Heavier companion AI use went with higher loneliness scores. Part of that link ran through displacement, as people who used the apps more were more likely to feel the apps had replaced real contact, and those who felt this way were lonelier. This route accounted for about a third of the overall association, and the rest of the link remained once it was taken into account.
The pattern depended on the person. Among those with moderate or high relational interdependence, the displacement route was clear and grew stronger as interdependence rose. Among those low in interdependence, it was not statistically significant.
People who define themselves through close ties showed a mixed picture. When they felt little displacement, they reported slightly less loneliness than others, but when they felt a lot, they reported markedly more. The author suggests that for these young adults, a sense that real relationships have been pushed aside may strike harder at their need to belong.
The study has clear limits. It was a one-off survey with no objective record of social time, so it cannot show that the apps cause loneliness or which came first. The questions on AI use and displacement were new and untested in another sample, and the displacement items asked people directly whether AI had cut into their social lives, which may have strengthened the link.
Other research has reported both benefits and risks from AI companions, and some studies find they can ease loneliness in the moment. The author suggests that efforts might focus on protecting time with friends and family while allowing moderate AI use, rather than on the apps alone.
