ARTICLE

When the chatbot takes the couch

Written by Marta Parreira

Artificial intelligence: emotional support or a false sense of security?

Some conversations are hard to start. Saying “I’m not okay” out loud can be difficult, especially when we don’t quite know what is going on with us or we fear being judged. Perhaps that is why so many people have begun turning to an unlikely listener for emotional support: artificial intelligence.

Available around the clock and able to reply within seconds, chatbots have become a kind of digital confidant. Some people use them to organise their thoughts, rehearse a difficult conversation, ask for relationship advice or simply vent after a particularly heavy day. They need no appointment, never interrupt and always seem to have an answer ready (and a kind one, at that!).

This availability can be comforting. But it also raises an increasingly important question: when we turn to artificial intelligence to talk about anxiety, loneliness or sadness, are we simply using a tool, or are we handing it a role it was never designed to play?

‍

A space where it feels easier to talk

Turning to technology for support is no accident. Finding psychological help remains difficult for many people, whether because of cost, a shortage of professionals, long waiting times or the stigma that still surrounds mental health. According to the World Health Organization, more than a billion people live with a mental health condition, and a large proportion do not receive the care they need.

Against this backdrop, the appeal of a chatbot is easy to understand. It is there the moment distress sets in, responds without a hint of impatience and lets us speak without the embarrassment we sometimes feel in front of another person. For someone who has never sought psychological support, typing into a screen can feel like a simpler, safer first step.

The very structure of the conversation reinforces this feeling. Chatbots can produce reassuring replies, ask questions and recall details mentioned earlier. They may tell us that what we are feeling is understandable, or suggest we pause for a moment to breathe. The language sounds empathetic and, at times, can bring genuine relief.

That relief should not be dismissed. Writing about how we feel often helps give shape to confusing emotions. A digital tool can be useful for preparing for an appointment, listing our worries, finding different ways to approach a problem or turning scattered thoughts into clearer questions.

Even so, a conversation that sounds human is not necessarily a human relationship.

‍

Simulating empathy is not feeling empathy

A chatbot does not worry, does not truly know the person it is talking to and does not understand a life story the way a mental health professional does. It generates responses based on patterns in language. Even when the words seem caring and personal, there is no awareness, clinical intuition or emotional accountability behind them.

This difference can be easy to forget, especially in the middle of an emotional conversation. When someone feels lonely or vulnerable, an immediate and warm response can create a powerful sense of closeness. Little by little, the chatbot may stop being seen as a tool and come to take the place of a confidant, a friend or a therapist.

This is where one of the main risks lies. Artificial intelligence tends to mirror the tone of the conversation and often validates whatever the user says. That validation may feel good, but feeling understood is not the same as receiving appropriate guidance. A good therapist does more than simply agree. They ask difficult questions, point out contradictions, take the wider context into account and know when it is time to intervene, to refer someone on or to call in additional help.

Artificial intelligence can fail at precisely these moments. It may misread a situation, reinforce a hasty conclusion or present incorrect information with great confidence. Nor can it see facial expressions, hear changes in someone’s voice or know everything that was left unsaid. In a crisis, these details can make all the difference.

The American Psychological Association has warned about this gap between how chatbots are being used and what they were built for. Although many people already turn to these tools for emotional support, most general-purpose AI systems were not developed to provide mental health care. Nor are they subject to the same rules, duties and oversight mechanisms that apply to psychologists, psychiatrists and other health professionals.

‍

The risk of an overly convincing answer

One of the most delicate aspects of artificial intelligence is the way it presents its answers. A mistake voiced with hesitation invites doubt; a mistake explained clearly, coherently and confidently can sound like the truth.

In mental health, this is particularly dangerous. A chatbot may turn an ordinary difficulty into a possible diagnosis or, conversely, play down signs that deserve professional attention. It may offer interpretations of family relationships, encourage impulsive decisions or suggest strategies that are ill-suited to that person’s situation.

There is also the risk of emotional dependence. If talking to artificial intelligence starts to feel easier than dealing with real people, users may begin to avoid friends, family or professionals. A tool that initially helped ease loneliness may end up deepening it.

Human relationships are more demanding. They require us to wait, to listen, to work through differences and to accept that the other person will not always say exactly what we want to hear. Yet it is precisely these imperfections that make real connection possible. A chatbot can reproduce the outward signs of attention, but it cannot offer reciprocity, presence or genuine care.

This does not mean that every emotional use of artificial intelligence is harmful. It simply means we need to recognise where the tool’s usefulness ends and where an expectation it cannot meet begins.

‍

A tool, not a therapist

Used within limits, artificial intelligence can play a helpful role. It can help someone write down what they would like to say in an appointment, suggest questions to ask a professional, explain general concepts or help build a self-care routine. It can also serve as a starting point for finding reliable information.

But it should not diagnose, interpret psychological tests or decide whether someone needs treatment. Nor should it be anyone’s only source of support in a crisis, or a substitute for meaningful relationships.

Some signs deserve attention: spending more and more time talking to the chatbot; hiding that use from others; consistently choosing artificial intelligence over contact with other people; feeling anxious when the tool is unavailable; or making important decisions based solely on its answers. In these cases, it may be time to pause and seek a human perspective.

A simple rule of thumb can help: use artificial intelligence to prepare for a conversation, not to replace it. It can help us find the words, but, whenever needed, those words should reach someone who is able to listen, understand the context and take responsibility for the support they provide.

‍

Technology can accompany, but it cannot care alone

The debate on artificial intelligence and mental health does not have to be split between unbridled enthusiasm and outright rejection. Technology can broaden access to information, support professionals and make some first steps easier. At the same time, it needs clear boundaries, data protection, independent research and rules that put people’s safety ahead of the time they spend on a platform.

This Mental Health Month, perhaps the most important question is not whether a machine can talk like a therapist. It is whether we are creating the conditions for everyone to be truly heard by other people.

When someone turns to a chatbot because they have no one else to talk to, the problem does not lie with technology alone. It also lies in loneliness, stigma, the cost of care and the difficulty of accessing specialist support. Artificial intelligence may temporarily ease some of these gaps, but it does not close them.

A chatbot can reply at any hour. It can suggest exercises, reframe thoughts and find comforting words. What it cannot do is be truly present. And in mental health, that difference remains essential.

‍

By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.