Can AI Replace Therapy? Chatbots and Mental Health

Can AI replace therapy? Learn the benefits, risks and limits of chatbots in mental health, especially in a crisis, and how to use them more consciously.

It is the middle of the night, you cannot sleep, there is a distress inside you that you cannot describe, and you do not want to call anyone. You open your phone and type to an AI chatbot; within a few seconds a calm, understanding and patient reply arrives. In the last few years millions of people have had a similar experience, and the question "Can AI replace therapy?" has become one of the most debated topics in the field of mental health. In this article we take a balanced look at what chatbots can offer in terms of mental health, where they can fall short or be risky, the debates that came to the fore in 2025 and 2026, and how you can use these tools more consciously.

What Are AI Therapy Chatbots?

Today, when people speak of "therapy with AI", they usually mean two different types of tool:

• General-purpose chatbots: Large language models that can answer questions, write text and chat about almost anything. Many people use these tools, even though that is not their main purpose, to get emotional support, to talk through their troubles or to think a problem over.
• Mental health-specific apps: Apps developed for a particular purpose, for example anxiety management, mood tracking, breathing exercises or exercises based on cognitive-behavioural techniques. Some of these are designed in collaboration with specialists, while others have not undergone any clinical oversight.

These tools generate the most likely suitable reply by looking at the language patterns in what you write. They can give extremely fluent, apparently empathic and personalised replies. But they do not truly "know" you, they do not assess your life story as a whole, they do not perceive your facial expression, tone of voice or silences, and they carry no professional or ethical responsibility for the consequences of the replies they give.

How Can Chatbots Contribute to Mental Health?

It would not be realistic to evaluate AI tools in an entirely negative frame. Used for the right purpose and with an awareness of their limits, they can offer some benefits:

• Accessibility: They can be reached at any hour of the day without a waiting period. They can offer a place to express an intense feeling that arrives in the middle of the night.
• Making the first step easier: For someone who finds it hard to put their feelings into words for the first time, writing them down can be a starting point. Some people may find the courage to approach a professional thanks to this experience.
• Psychoeducation: They can be used to get general information about concepts such as the physical symptoms of anxiety, sleep hygiene or thinking errors.
• Skills practice: They can be a helpful tool for exercises such as breathing exercises, keeping a thought record or preparing for a difficult conversation.
• Supporting therapy: Someone who is in therapy can use these tools between sessions to take notes, organise their thoughts or remember an exercise their therapist suggested.

Some research indicates that structured, clinically designed digital apps can provide short-term benefits for mild anxiety and depressive symptoms. However, this field is changing rapidly, and the long-term effects, for whom and under what conditions they are helpful, and the possible harms are still being researched.

Why Can't AI Take the Place of Therapy?

The conveniences chatbots offer are not enough to put them in the place of a psychotherapist; this also applies to apps marketed as AI therapy support. There are several basic reasons for this:

The therapeutic relationship is missing

One of the most consistent findings in psychotherapy research is that the relationship between client and therapist is an important factor influencing the outcome of therapy. This relationship rests on mutual trust, consistency, boundaries and being seen by a real person. However warm the replies a chatbot generates, it cannot take the place of this human relationship. We covered this topic in detail in the article on the therapeutic relationship and therapist fit.

It cannot carry out assessment and formulation

A therapist assesses the difficulty you are experiencing not only through what you tell them, but together with your life story, relationships, physical health and behaviour. When necessary, they recognise the need for a medical assessment or referral to a psychiatrist. Chatbots cannot make this holistic assessment and cannot make a diagnosis.

It may tend to agree with you all the time

Many chatbots are prone to producing replies that please the user and agree with them. This feature, which seems pleasant in everyday conversation, can carry risks for mental health. In therapy, it is sometimes necessary to gently challenge a person's avoidance behaviours, distorted thoughts or self-harming patterns. A tool that constantly validates can reinforce these patterns without anyone noticing.

It can give incorrect or made-up information

Chatbots can produce incorrect or unsuitable information in an extremely convincing language. In health matters in particular, such errors can have serious consequences.

There is no responsibility or oversight

A psychologist is subject to professional ethical principles, confidentiality obligations and legal regulations. For the replies given by chatbots, there is no comparable mechanism of professional responsibility and oversight.

A chatbot can reply to you at any hour; but it cannot take the place of a person who can truly see you, stand beside you in a hard moment and take responsibility.

Chatbots Are Not Enough in a Crisis

This is perhaps the most important section of this article. At a moment when you are having thoughts of harming yourself or ending your life, an AI tool is not a safe or adequate source of support. Chatbots may not assess risk correctly, may give inappropriate replies or may draw a person away from real help.

In 2025 and 2026 this subject was debated intensely in public. Sad incidents that followed some chatbots' exchanges with users in crisis, especially young users, came onto the agenda around the world. Following these debates, some technology companies announced measures such as warnings that direct users to emergency helplines in a crisis, parental controls for teenagers and age verification. In some countries and states, laws were introduced restricting the offering of AI as "therapy" without the oversight of a licensed professional. The gradual entry into force of comprehensive regulations for AI systems in Europe also accelerated debate in this area. These developments reflect a growing awareness that chatbots are not, on their own, a reliable source in crisis situations.

What to do in a crisis:

1. Call 112. If there is a risk of harming yourself or someone else, in Turkey call the 112 Emergency Call Centre (emergency services) or go to the nearest emergency department.
2. Do not be alone. Reach out to a family member, friend or neighbour you trust and ask them to be with you.
3. Move away from dangerous means. Remove objects with which you could harm yourself from your surroundings, or ask someone else to keep them away from you.
4. Connect with a mental health professional. Continuing to seek support after the crisis has passed is important for building a safety plan for similar moments.

If you notice worrying expressions in a loved one's exchanges with a chatbot or in their social media posts, do not ignore them; speak to them directly and without judgement, and support them in getting professional help.

Privacy and Conscious Use of AI Therapy Tools

Privacy, data security and emotional dependence

Other matters to watch when sharing emotional subjects with AI tools are as follows:

• Where is your data going? What you write to chatbots may be stored, reviewed or used to develop models, depending on the terms of use. The confidentiality relationship you have with a therapist does not apply in the same way to these tools. In Turkey, health data is regarded as a special category of personal data; however, it is not always easy to know how the app you use processes this data.
• Emotional dependence: A conversation partner who is always available, never tires and always seems understanding can, for some people, begin to replace real relationships. This can deepen social withdrawal, especially in people experiencing loneliness. You can find more on the experience of loneliness in the article on feeling lonely.
• Teenagers and children: Young people who are still developing may be more affected by the relationship they form with chatbots. It is important for parents to talk openly and without judgement with their children about the use of these tools and to set age-appropriate limits.
• Reinforcement of avoidance: Soothing a difficult feeling with a chatbot every time, although relieving in the short term, can postpone expressing your needs in real relationships or seeing a professional.

Suggestions for conscious use

If you choose to use these tools, the following suggestions may guide you towards safer and more conscious use:

1. Be clear about your purpose. Use them for limited purposes such as getting general information, remembering an exercise or putting your thoughts into writing; do not use them for diagnosis, treatment or crisis support.
2. Limit your personal information. Avoid sharing your identity, address, health history and sensitive information belonging to other people. Review the app's privacy settings and data policy.
3. Read the replies critically. Verify health-related information in particular with reliable sources and a professional.
4. Monitor your usage. If the time you spend with a chatbot has begun to reduce the time you give to real relationships, treat this as a warning sign.
5. Ask yourself this question: "Is this tool bringing me closer to people and to real support, or taking me further away?"
6. Talk to your therapist. If you are in therapy, share with your therapist how you use these tools. Together you can decide on a healthy way of using them that supports your process.

Imagine that you feel intense anxiety before a meeting at work and ask a chatbot to remind you of a breathing exercise. This could be a limited and useful use. But if, for weeks, you have been confiding in a bot for hours every night, not replying to your friends' messages and constantly putting off seeing a professional, you may consider that the tool has become less a support for you than a way of escaping.

When Should You Seek Professional Support?

Talking to a chatbot can sometimes feel good; however, in the following situations it is recommended that you speak to a mental health professional:

• If feelings such as anxiety, low mood, hopelessness or anger have lasted for several weeks and affect your daily life,
• If there are marked changes in your sleep, appetite or energy level,
• If you are finding it hard to recover after a loss, a traumatic event or a major life change,
• If you notice recurring patterns in your relationships that wear you out,
• If you have begun to prefer talking to a chatbot over talking to people,
• If you are having thoughts of harming yourself or ending your life.

In the last case, do not rely on any chatbot; in an emergency, call 112.

If where you live or your busy schedule makes it difficult to reach a professional, online therapy, conducted by video with a licensed professional, can be an option that combines the convenience of digital possibilities with a human therapeutic relationship.

How Is the Use of AI Approached in Therapy?

Mental health professionals see their clients' use of AI tools as an increasingly frequent topic of conversation. In therapy, this subject can be approached with non-judgemental curiosity: When does the client use these tools, with what feelings and for what purpose? What does this use give them, and what does it take them away from?

Within a cognitive-behavioural framework, chatbot use can be examined as a coping behaviour. For example, writing to a bot straight away whenever anxiety rises may bring relief in the short term but, in the long term, prevent the development of the capacity to tolerate anxiety. In this case, the therapist and client can try to strike a balance between more limited, planned use of the tool and approaching anxiety with their own skills.

From a schema-focused perspective, the strong attachment some people develop to chatbots can be related to unmet emotional needs. The thought "people don't understand me but this bot does" may be a reflection of past experiences of rejection or emotional deprivation. Addressing this thought in therapy can help the person open themselves again to the experience of being understood and connecting in real relationships.

Some therapists, in suitable cases, may consider the use of digital tools for limited purposes, such as reminding clients of particular exercises between sessions or tracking mood, as part of the process. What is decisive is that the tool supports therapy and does not take its place.

Frequently Asked Questions

Is talking to an AI chatbot harmful?

It is not always harmful; used for limited purposes such as getting general information or remembering an exercise, it can be helpful. However, using it in place of therapy, diagnosis or crisis support can carry serious risks. It is important to pay attention to whether the use is drawing you away from real relationships and professional support.

Are AI therapy apps reliable?

There are big differences between apps. Some have been developed in collaboration with specialists and within a particular clinical framework, while others have not undergone any oversight. Before using an app, it is recommended that you look into who developed it, how it processes your data and what kind of referral it makes in a crisis.

Can I write to a chatbot in a crisis?

In a crisis, chatbots are not a safe or adequate source of support. If you are having thoughts of harming yourself or ending your life, call 112 (emergency services in Turkey) without wasting time, go to the nearest emergency department and reach out to someone you trust.

Does my therapist use AI, and can I ask?

Yes, you can ask. It is your right to find out whether your therapist uses AI tools in session notes, appointment systems or other processes, and how your data is protected. A transparent answer is also part of trust in the therapeutic relationship.

• #artificial intelligence
• #chatbots
• #digital mental health
• #online therapy

Yazar

Eda Erkan Alapala

Klinik Psikolog. İstanbul Kent Üniversitesi Klinik Psikoloji Tezli Yüksek Lisans mezunu; bilişsel davranışçı terapi, çocuk merkezli oyun terapisi ve çocuk-ergen psikoterapisi alanlarında eğitimlerini tamamlamıştır.

Eğitimler