Can ChatGPT replace your therapist? Mental health experts warn of hidden risks
AI may be able to answer at 2am. It may even sound like it understands. But mental health professionals warn that when someone is anxious, depressed, traumatised or in crisis, sounding human is not the same as being human.
That is the concern growing around the use of artificial intelligence (AI) chatbots for mental health support, as more people turn to tools such as ChatGPT for answers about stress, anxiety, depression and emotional distress.
In the UK, “Mental Health UK” reported in November 2025 that 37% of adults had used AI chatbots to support their mental health or well-being.
But mental health professionals say that convenience can become dangerous when AI is treated as a therapist, psychiatrist or crisis counsellor.
“The greatest risk of turning to AI for mental health support is bad advice potentially with tragic outcomes,” said Professor Christopher Szabo, a member of the South African Society of Psychiatrists (SASOP).
He said that while AI can be useful in mental healthcare, including administrative work, screening, assessments and analysing patient data, it cannot replace the one-on-one relationship between a patient and a trained human professional.
“AI chatbots can provide inaccurate or misleading information and false reassurance and reinforce harmful thinking,” he said. Because an AI is designed to always provide an answer, the chatbots have been known to ‘hallucinate’ or make up answers.”
That matters because a person in emotional distress may not always have the ability to question whether the answer they have received is accurate.
Dr Melané Van Zyl, a SASOP board member told us that AI cannot reproduce the clinical judgement, personal awareness and therapeutic relationship that human mental health professionals provide.
“AI has several limitations. Firstly, AI cannot fully understand human context and complexity. The responses are generated from learned data and predictable language patterns rather than a genuine understanding of a person’s experience. It cannot observe important nonverbal signals, including body language and facial expressions and subtle changes in language,” said Van Zyl.
The difference, she explained, is not simply about whether a chatbot can offer comforting words.
A human therapist can notice when a patient’s voice changes, when they avoid eye contact, when their behaviour shifts or when what they are saying does not match how they appear.
A clinician can understand the patient’s personal history, relationships, culture and circumstances and use that context to assess risk and decide what needs to happen next.
“That human connection is not a small part of therapy. It is central to it.”
“Therapy is fundamentally more than just the delivery of information or advice,” Van Zyl said.
It involves understanding a person’s lived experience, interpreting verbal and nonverbal communication, recognising changes in mood and behaviour, challenging harmful assumptions appropriately, establishing trust, making an accurate diagnosis and responding safely during a crisis.
AI, she said, cannot take on the ethical, professional and legal responsibility for a person’s care.
There are also concerns about privacy. Unlike a consultation with a registered mental healthcare professional, conversations with an AI chatbot may not have the same confidentiality protections or professional oversight.
And then there is the danger of dependency.
An AI chatbot may respond immediately, never appear tired and always be available. For someone who feels isolated, that can create the impression of a deeply personal relationship.
But the chatbot does not have a lived experience, emotions or genuine understanding of the person on the other side of the screen.
“It is not only about diagnosis of symptoms and prescription of treatment, but the quality of the interpersonal connection, the relationship developed between doctor and patient, that drives the positive impact of mental health treatment on a patient’s quality of life,” Szabo said.
The message from mental health professionals is not that AI has no place in healthcare.
Wellness apps, mindfulness tools and carefully designed digital technologies may support people with everyday well-being and help professionals work more efficiently.
But they should supplement, not replace, counselling, psychotherapy, psychiatric assessment or clinician-led care.
The reality is that AI is already part of people’s lives. The question is no longer whether people will use it for mental health support. It is whether they will know when a chatbot is no longer enough.
“If someone finds themselves turning to AI because they are experiencing mental distress or struggling to cope with daily life, it is important to seek professional help,” Szabo said.
“They should contact a mental health professional or a helpline rather than relying on AI alone.”
For anyone in South Africa experiencing severe emotional distress, thoughts of self-harm or feeling unable to cope, reaching out to a qualified mental healthcare professional or a crisis support service such as the South African Depression and Anxiety Group (SADAG) is crucial.
More articles for you...
Important Disclaimer: The articles shared on this blog come from reputable sources and are intended solely for informational purposes. They do not represent the opinions or views of Yes, We Care / The WeCare Independent Pharmacy Network. While we strive to provide accurate and reliable information, we cannot guarantee its completeness or relevance. Readers are encouraged to conduct independent research and seek professional guidance when necessary. Yes, We Care / The WeCare Independent Pharmacy Network assumes no responsibility for decisions made based on the content provided.