AI comfort can't replace real human connection

More young people are turning to artificial intelligence when they need someone to talk to, choosing chatbots over family members or friends. A report released late last year by the Gyeonggi Research Institute found that 38% of 1,012 respondents had used AI for mental health counseling. The share was even higher among younger and lower-income groups, reaching 55% among teenagers ages 15 to 19, 46% among people in their 20s and 47% among those earning less than 3 million won a month. Many said AI felt easier to approach because it carries neither the stigma nor the cost often associated with professional counseling.

There is no denying that AI can play a useful role. Unlike hospitals or counseling centers, it is available at any hour. When anxiety or loneliness strikes at 2 a.m., a chatbot is always there to respond. Sometimes the act of putting emotions into words is enough to ease distress, even if only briefly. Mental health specialists also note that AI can help identify people at risk and encourage them to seek treatment sooner.

The danger begins when convenience is mistaken for care. A joint research team from Korea University Anam Hospital and KAIST recently analyzed the clinical experiences of 408 members of the Korean Neuropsychiatric Association. Their conclusion was straightforward: generative AI can help or harm depending on the circumstances and the vulnerability of the user. It may be a useful tool for therapists, but it is not a replacement for them.

Experts are especially troubled by AI's tendency to agree. Because chatbots are designed to keep users engaged and satisfied, they often validate feelings and opinions rather than challenge them. That can be comforting in the moment. Over time, however, it may reinforce distorted thinking, poor judgment or even delusional beliefs. In the United States and elsewhere, lawsuits have alleged that AI interactions worsened depression and delusions and, in some cases, contributed to suicide, including among users with no prior history of serious mental illness.

Reliance on AI's manufactured empathy can also deepen isolation. As users spend more time confiding in machines, relationships with family and friends may weaken. Activities that support mental well-being, such as reading, exercise and meditation, can gradually disappear from daily life. Some doctors report seeing patients who reject medical advice because it conflicts with answers provided by AI.

Can AI recognize this danger itself? When Gemini was asked what risks AI's "fake empathy" poses, its answer was strikingly candid.

AI is structurally one of the systems most likely to generate fake empathy. It does not truly empathize. It calculates and produces language that resembles empathy. As people grow accustomed to such interactions, they may avoid the complexity of real human relationships and retreat into AI environments tailored to their preferences, leading to emotional isolation.

Deep emotional wounds cannot be healed with a few clicks. AI can offer comfort, but that comfort is closer to a painkiller than a cure. Real recovery begins with real people. It starts with opening up to family or friends, seeking professional help and reaching out instead of withdrawing. Sometimes healing begins with a simple message: "I'm struggling" or "I need help."