AI companions make lonely people even lonelier over time, research shows
Researchers say the fundamental flaw of AI companions lies in their structure, not their level of technology
One in five Americans have used an artificial intelligence tool specifically designed to simulate a romantic partner, whilst 72% of teenagers have tried an AI companion at least once, according to a 2024 survey by Common Sense Media.
Yet despite the rapid uptake of such technology, a growing body of research suggests that turning to AI for emotional connection may be doing more harm than good.
Year-long study reveals a troubling pattern
A study conducted and published in Psychological Science tracked over 2,000 people across 12 months and found that those who relied on artificial intelligence for social interaction became lonelier over time, not less so.
Scientists monitored chatbot usage and levels of loneliness amongst adults from English-speaking countries for a full year, taking into account life-changing factors such as relationship breakdowns or relocation that might otherwise skew the results.
The overall conclusion pointed to a clear and consistent trend: those who were already more lonely and who increased their use of chatbots over the course of the year experienced a measurable rise in loneliness.
Human contact outperforms AI by a significant margin
A separate study from the University of British Columbia put AI companions to a direct test, focusing on university first-year students — one of the most socially vulnerable groups at any given time.
Students who sent daily text messages to a randomly assigned fellow first-semester student over two weeks reported approximately a 9% reduction in loneliness.
By contrast, students who messaged a Discord chatbot daily over the same period saw only a 2% reduction — a figure statistically equivalent to writing a single sentence in a journal each day.
The gap between human contact and AI contact was, by any measure, considerable.
Fundamental flaw lies in the structure of AI, not the technology
According to Ruo-Ning Li, the lead author of the UBC study, the core problem with AI companions is not a matter of technological limitation but of structural design.
"When you're talking with a chatbot, you can get a lot from it, but you never have the chance to give something back," said Li during a discussion amongst scientists.
Li argues that genuine human connection depends not only on receiving support and feeling cared for, but equally on being able to give help to another person.
Organisational psychologist Adam Grant echoed this view, stating that meaningful relationships are built on the reciprocal provision of support — something AI chatbots are fundamentally incapable of offering. In his assessment, AI can never function as a true companion, only as a servant.