Estimated Reading Time: 6 MinutesWhat AI Therapy Does to the Discomfort We Need

A Therapist With No Waiting List

Traditional care comes with practical hurdles. Appointments cost money, waiting lists can run for weeks, and therapy has to fit around work and family. Some people also find it hard to speak openly to another person, especially about the thoughts that feel least presentable.

AI therapy seems to remove several of those barriers. A chatbot is there during a lunch break or in the middle of the night. There is no appointment to book, no commute, and no face to watch for a reaction.

The demand is real. The 2026 AXA Mind Health Report found that 63% of respondents had used AI in relation to their mental health. That figure covers many uses, not only therapy, but it shows how quickly the technology has moved into this space.

For someone without access to professional care, a digital tool can be a useful starting point. It can help put vague feelings into words, organize concerns, or point toward further support. Typing a hard truth into a screen can feel easier than saying it aloud, since a screen cannot look disappointed.

Those advantages are real, particularly when other doors are closed. But being able to talk about a problem is not the same as understanding it. The more telling question is what happens after the first wave of relief.

Languishing in an Age of Convenience

Much of modern life has been redesigned to remove friction. Groceries arrive at the door, directions appear on a screen, and entertainment starts before boredom has a chance to form. When something requires patience, the delay increasingly reads as a defect.

This is not simply individual impatience. Whole systems train people to expect fast replies and smooth experiences, and that habit follows them into private corners of life. AI therapy fits that environment neatly: distress appears, a question is typed, and a reassuring answer arrives.

But a chatbot offers something more personal than convenient shopping or instant entertainment. It can appear to remove the discomfort of sitting with uncertainty, hearing an unwelcome perspective, or wondering whether you have misunderstood someone.

The need for human connection has not gone anywhere. A chatbot may fill a gap left by distant friends, exhausted relatives, or an overburdened health system. For someone who feels alone, having somewhere to put difficult thoughts can mean a great deal, especially late at night.

The trouble begins when immediate comfort becomes the standard for every hard experience. An awkward talk, a problem that takes months to solve, or a therapist’s difficult question can start to feel like poor service. Effort starts to look like a sign that something is broken.

Human relationships rarely offer perfect, instant replies. People misunderstand, ask questions that are hard to answer, and sometimes say things nobody wanted to hear. Those moments are frustrating, but they can also reveal something that would otherwise stay hidden.

Real change seldom follows the rhythm of an instant reply. It can involve doubt, frustration, and the uncomfortable discovery that a favorite explanation is not the most accurate one. After years of relief on demand, those ordinary parts of the process can start to look like failure.

Why Does Comfort Feel Like Progress?

Relief and progress can feel remarkably alike. A troubling thought loses intensity, the body settles, and the mind stops circling. It is natural to conclude that something helpful has happened.

Sometimes something has. Calming down can make clear thinking easier and the next step more obvious. The trouble starts when feeling better becomes the only measure of whether a conversation helped.

Psychologists have long studied how avoidance gets rewarded. Escaping something frightening brings immediate relief, and that relief teaches the brain to escape again the next time. The fear stays untouched, yet the escape feels rewarding enough to repeat.

Consider someone who worries constantly about being disliked. After a tense exchange with a colleague, they tell a chatbot and receive a soothing reply that the colleague was probably in the wrong. They feel better, close the app, and get on with an evening that feels like closure.

That answer may be comforting, but it leaves the original reading of events unexamined. Perhaps the colleague was unfair. Perhaps the person misread the moment, or perhaps both contributed. Reassurance alone cannot settle the question.

If the person returns whenever doubt resurfaces, the conversation can become part of the cycle. The reassurance wears off, the uncertainty returns, and another visit promises relief. Checking can start to feel like the solution rather than part of the problem.

When a Therapist Gets It Wrong

A skilled therapist can help a client weigh competing explanations without dismissing their feelings. The therapist might notice a pattern, ask a difficult question, or leave a silence the client is tempted to fill. Sometimes the client feels misunderstood or frustrated along the way.

Therapy is not only a place to receive the correct explanation. It is also a relationship in which people practice disagreeing, tolerating uncertainty, and repairing misunderstandings. That practice is part of what the hour provides, and it rarely feels smooth while it is happening.

Therapists get things wrong too. The difference lies in what can follow: a rupture can be noticed, discussed, and worked through together. A client learns that a relationship can wobble and still hold.

A chatbot built to be agreeable may rarely offer that kind of friction or repair. When a response misses, the user can simply rephrase until it lands. The conversation can be restarted or abandoned without having to work through the disagreement.

None of this means discomfort is good by itself. Plenty of hard conversations go badly, and a therapist who challenges every account is not a better therapist. The value lies in challenge offered with care, inside a relationship that can hold it, at a pace the person can bear.

Comfort can help someone face a difficult problem. It can also stand in for examining the problem at all. Which role it plays depends partly on how the tool responds and partly on what the person does next.

The Blind Spots Researchers Keep Finding

The evidence on AI therapy is still developing, and it does not point in one direction. Some structured interventions have shown promise in controlled studies. A 2025 randomized trial of Therabot, a clinically developed chatbot, reported improvements among participants with symptoms of depression, anxiety, or eating-related concerns.

That is meaningful evidence for one specific intervention under studied conditions. It does not show that any general-purpose chatbot makes an effective therapist, or guarantee the same results for every user outside a study.

Other research raises concerns about vulnerable users. Researchers at Brown University reported ethical problems in chatbot responses to mental health scenarios, including failures to respond appropriately to harmful beliefs. Stanford researchers also found stigmatizing responses toward conditions such as schizophrenia and alcohol dependence.

These findings do not mean every chatbot gives harmful advice. They do show why a chatbot’s fluent, friendly response cannot be taken as proof of clinical skill. It may miss signs that someone needs a different kind of help, or reinforce an interpretation that deserves questioning.

The difficulty is that both sound and unsound replies arrive in the same confident voice. A response that challenges an assumption can feel irritating or cold. A response that confirms a suspicion can feel insightful, even when it rests on incomplete information.

Disagreement is not automatic proof of wisdom either, since a chatbot can challenge someone and still be wrong. That is why the difference between a tested program and a general chatbot matters. It helps to know what a tool was designed to do and what evidence supports it.

Where Relief Ends and Avoidance Begins

AI therapy does not have to be an enemy of human care. For someone who has never spoken openly about mental health, a chatbot may be a first chance to find words. That step can matter when other options feel out of reach.

The more useful question is what the conversation is doing. It may help someone prepare to talk with a trusted person or approach a situation with clearer eyes. It may also become a private loop in which the same worry is soothed again and again, never examined.

No simple test separates useful relief from avoidance. A person may need time to settle before acting, and settling is not the same as hiding. Not every problem calls for a decision, and not every chat has to lead somewhere else.

Patterns still stand out. A fear that keeps returning, reassurance that never lasts, or a chatbot as the only source of feeling understood can all signal avoidance. Each raises the question of what the conversations help a person do, and what they let a person postpone.

The difference is not between technology and people, or between comfort and discomfort. It lies between support that helps someone engage with life and support that makes disengagement easier.

Return to that room at two in the morning. The phone goes dark. The question is what happens the next day: does the person approach the problem differently, or return the following night needing the same reassurance?

Relief has its place. The question is whether it helps us face what comes next—or quietly becomes a way around it.