by Psychological Science research finds lonely people turn to AI companions, and heavier use was followed by greater emotional isolation
Millions of people now talk to AI chatbots for company. Is that a comfort for loneliness or a contributor to it? A study in Psychological Science, by Dunigan Folk and Elizabeth Dunn and covered by PsyPost in May 2026, tracked more than two thousand adults for a year and found a pattern that cuts both ways. It does not settle the question, and the authors say so, but it is the kind of longitudinal evidence that the debate has needed.
The Design
The paper, titled How Does Turning to AI for Companionship Predict Loneliness and Vice Versa, followed 2,149 adults: 50 percent from the UK, 28 percent from the US, 14 percent from Canada and 8 percent from Australia. Of these, 979 completed all four survey waves and 466 completed three. Participants averaged about 40 years old, and 49 percent were men. Over 12 months, they reported how often they used chatbots for social purposes, such as seeking advice, having regular social conversations or finding companionship, and rated their emotional isolation and broader social connection. Major life events, including moving, breakups, starting a steady relationship or becoming a parent, were tracked as controls.
How Common Was Social Chatbot Use
Roughly 26 to 30 percent of participants used chatbots for social purposes at any given wave, and the average frequency did not change significantly over the study period. That is a large share, and it was stable. The study therefore describes habits that were already established, not a short-lived novelty.
What They Found
Higher emotional isolation at one time point predicted greater social chatbot use four months later. In other words, loneliness drives people toward chatbots. Greater chatbot use was then followed by further increases in emotional isolation at the next time point, suggesting that the reverse effect may also occur. A different measure, overall social connection, showed a different picture: lower social connection predicted increased chatbot use later, but chatbot use did not predict lower social connection. Major life events were not linked to later chatbot use, and a breakup was followed by reduced social connection but not by increased chatbot use.
The authors suggest that easy, always-available AI interactions may be crowding out more rewarding relationships with humans. They describe the evidence as initial and urge caution, writing that they urge caution in drawing strong conclusions given the exploratory nature of their analyses.
Limits to Keep in Mind
The study is observational, so it cannot establish that chatbot use causes anything. All the data were self-reported, which is vulnerable to bias. The analyses were exploratory. The sample came from four English-speaking countries and may not reflect other cultures. And the measure of emotional isolation differs from the measure of social connection, so results that look contradictory may reflect different aspects of loneliness.
The Wider Debate
The paper sits in a contested field. The article notes that over a billion people reportedly use large language models and generative AI tools by 2026. Some researchers see AI companionship as a scalable response to loneliness. Others warn that it may cause harm, partly because AI cannot genuinely reciprocate emotional self-disclosure. A commentary in Psychiatric Times cites a 2025 study finding that chatbots can reduce loneliness because users feel heard, alongside a 2025 OpenAI and MIT Media Lab study linking heavy use with increased loneliness, possibly because it displaces human contact. Both can be true for different people and different patterns of use.
A Useful Distinction
The difference may lie in how the tool is used. Occasional use to rehearse a difficult conversation, draft a message or practise a language is not the same as relying on a chatbot as the main source of emotional support. The first adds to a person’s social world, and the second may substitute for it. The question to ask is whether the interaction leads back toward people or away from them.
Practical Guidance
For individuals, the evidence supports a few modest steps. Notice how you feel after using a companion app: if you feel better and more able to approach people, it may be helping; if you feel more withdrawn, it may not be. Keep some regular human contact, even if it is brief. Be careful about sharing private details with commercial apps, since privacy protections vary. For parents, ask young people how they use these tools, and keep the conversation open and curious.
For clinicians, it is reasonable to ask patients about emotional reliance on AI companions, particularly when loneliness, grief or social withdrawal is present. For companies, the lesson is that design choices which maximise engagement may not maximise wellbeing.
What Researchers Still Need to Learn
Future work should test whether particular designs make a difference, whether effects differ by age, gender and personality, and what happens over longer periods. Experimental studies, in which some participants are randomly given access to companions, could shed light on cause and effect. Until then, the cautious reading is the right one.
Our London pages have kept a close eye on the subject, including coverage of claims that chatbots make Britain stupider, UK plans to regulate them and a man who assigned his chatbot a persona. See also our guide to London news coverage, and for American coverage Bohiney Magazine.
For some lighter reading, The Daily Mash is just a click away.
SOURCE: bohiney.com
