Turning to chatbots when lonely may exacerbate feelings of loneliness, study finds

A 12-month longitudinal study of adults from four English-speaking countries found that being lonely may spur people to seek companionship through chatbots. However, such use may, over time, exacerbate feelings of loneliness. The paper was published in Psychological Science.

The recent development of large language models, AI systems capable of communicating with users in natural human language, has ushered a new era for humanity. It can be argued that it has started both a revolution in the way business is conducted and in how people live their lives. Soon after its release in late 2022, ChatGPT gained millions of users. Four years later, in 2026, estimates state that over 1 billion people use large language models and generative AI tools.

In this short period, many people have started using AI chatbots for companionship. Some researchers suggest that AI companionship might prove to be a scalable tool for combating the loneliness epidemic. However, there are also voices warning that AI companions might do more harm than good. The main reason for this is that AI is unable to genuinely experience emotions or bond with another person; it can only supply an illusion of understanding, empathy, caring, and love.

Furthermore, because AI lacks a genuine inner life, it cannot engage in the reciprocal self-disclosure that is critical for building rewarding, long-term human relationships. Consequently, forming bonds with AI companions or even AI romantic partners might ultimately leave vulnerable individuals unsatisfied and worsen their mental health over time.

Study authors Dunigan Folk and Elizabeth Dunn note that people may turn to AI companions because they are always available (unlike human partners or friends) and simulate human companionship in a manner that feels compelling in the moment. However, the researchers theorize that these easy but shallow interactions with AI may be crowding out the more rewarding interactions with real humans in people’s lives.

To examine this, they conducted a 12-month longitudinal survey investigating the relationship between loneliness and the use of chatbots for social purposes. The study participants included 2,149 individuals from the United Kingdom (50%), the United States (28%), Canada (14%), and Australia (8%) who completed at least one survey across four data collection waves. Of these participants, 979 completed all four surveys, and 466 completed three. The participants’ average age was 40 years, and 49% were men.

The surveys asked participants how often they had used chatbots for social purposes over the past four months—such as asking for advice on life decisions, having regular social conversations, or seeking social companionship. The surveys also measured “emotional isolation” by asking participants how emotionally isolated they felt from other people during that same period.

dditionally, the researchers tracked a broader metric of overall “social connection” and asked whether participants had experienced major life events—such as a relocation, a breakup, the beginning of a steady romantic relationship, or becoming a parent—to control for outside social stressors.

Results showed that roughly 26% to 30% of participants reported using chatbots for social purposes at any given wave. On average, the frequency with which participants used chatbots for these purposes did not change significantly across the data collection period.

Crucially, individuals who reported feeling more emotionally isolated at one time point tended to report a higher use of chatbots for social purposes four months later. This confirmed the study authors’ expectations that feeling lonely leads people to seek out chatbots. However, after increasing their chatbot use, participants also tended to report even further increased emotional isolation at the next time point. Experiencing a major life event, on the other hand, was not associated with a subsequent increase in chatbot use.

When the study authors looked at their broader measure of overall social connection, the results showed that individuals who felt less socially connected also tended to increase their social chatbot use four months later. In this analysis, however, increased chatbot use did not lead to even lower social connection at a later time point. (Experiencing a breakup tended to be followed by reduced feelings of social connection four months later, but it was not associated with increased chatbot use. Other types of life events were not associated with changes in social connection.)

“Taken together, these findings provide initial evidence that being lonely may spur people to seek companionship through chatbots but that such use may, over time, exacerbate feelings of loneliness. We urge caution, however, in drawing strong conclusions given the exploratory nature of our analyses,” the study authors concluded.

These findings contribute to emerging evidence about the ways people use AI chatbots to satisfy psychological needs. However, it should be noted that the observational design of this study does not allow for definitive causal inferences to be derived from the results. Additionally, all findings were based on self-reports, leaving room for reporting bias to affect the results.

The paper, “How Does Turning to AI for Companionship Predict Loneliness and Vice Versa?” was authored by Dunigan Folk and Elizabeth Dunn.