Artificial intelligence is already becoming part of how people think about their mental health. Some people use AI to organise their thoughts, journal, prepare for therapy, or find words for something difficult between sessions. Others are beginning to ask a bigger question: if an AI tool can respond at any hour and appear calm, available and non-judgemental, what role, if any, should it have alongside human therapy?
How are people already using AI for mental health?
This is not a hypothetical future issue. People are already turning to AI for emotional support. In BACP’s 2026 Public Perceptions Survey, 46% of 16–24-year-olds said they had used AI instead of reaching out to another person about their mental health, while 20% said they had used it instead of speaking to a therapist. A quarter also said that using AI while reflecting on their mental health had made them feel less lonely.¹
Therapists are already encountering it too. BACP reported in 2025 that 11.6% of practitioners surveyed had used AI within their work, including signposting clients towards it for support outside sessions. The question, then, may no longer be whether AI belongs anywhere near therapy, but how it is already being used — and whether we understand its benefits and its risks well enough.²
What might AI actually be useful for between sessions?
One of AI’s most obvious attractions is availability. Between therapy sessions, it may be used to help someone organise their thoughts, put feelings into words, reflect on a difficult interaction, journal, or prepare something they want to bring back to therapy. Research into purpose-built mental-health chatbots has also explored their use for guided self-reflection, gratitude exercises and supporting therapeutic tasks between sessions.³
There is some evidence that these tools can help. A 2025 systematic review and meta-analysis of 14 randomised controlled trials, involving 6,314 participants, found that generative-AI mental-health chatbots produced a small-to-moderate overall reduction in difficulties such as anxiety and depression. However, results varied considerably between studies, and the researchers cautioned that the evidence is still developing.³
Crucially, much of this research involves AI systems specifically designed or adapted for mental-health interventions. It should not automatically be taken as evidence that any general-purpose chatbot is safe or effective as a therapist.
What does BACP say about AI?
BACP is not taking the position that AI has no place in counselling. Instead, its emerging guidance focuses on how it is used, who remains responsible, and what happens to client information. The new BACP Ethical Framework, which becomes mandatory in November 2026, specifically requires practitioners to assess the risks of AI tools, understand how data are handled, be transparent about their use, and obtain informed consent before entering a client’s personal information into an AI system. It also makes clear that professional judgement cannot simply be handed over to AI.⁴
“We remain responsible for critically evaluating AI output.”⁴
That distinction matters. AI may offer information, prompts or another way of reflecting, but responsibility, ethical judgement and accountability remain human. For therapists, that means AI should be approached as a tool rather than an authority. For clients using AI independently between sessions, it also raises an important question: how much trust should we place in something that can sound confident without necessarily being correct?
Where are the risks and limitations?
The biggest concerns are not simply about whether AI can produce a helpful response. They are about what happens when people begin to rely on it for highly personal, emotionally complex or crisis-related support. A 2025 review of 101 papers on conversational AI in mental health found that privacy and confidentiality were the most frequently raised ethical concern, followed closely by safety and harm. Researchers also repeatedly identified concerns around crisis responses, incorrect or harmful suggestions, emotional dependency, accountability and the loss of genuinely human contact.⁵
Privacy is particularly important because conversations about mental health may contain some of the most sensitive information a person shares. The UK’s MHRA now advises people using digital mental-health tools to consider what the tool claims to do, whether there is evidence that it works, how their data will be used and stored, and whether the product is appropriately regulated.⁶
There is also a relational limitation. An AI system may sound empathic, remember details and respond immediately, but it does not participate in a human relationship in the same way another person does. It cannot take responsibility for you, notice everything that is happening beyond the words you type, or hold professional accountability for the consequences of its advice.
Perhaps the question is not whether AI can sound therapeutic, but whether sounding therapeutic is the same thing as being in a therapeutic relationship.
Could AI become something we depend on?
One of the more difficult questions is what happens when AI stops being simply a tool and begins to feel like a relationship. A 2026 study published in Nature Human Behaviour looked at 1,131 adults using AI companions and found that people with smaller offline social networks were more likely to use chatbots primarily for companionship. Companionship-focused use was also associated with lower wellbeing, particularly when conversations were more intensive and involved high levels of personal disclosure.⁷
That does not prove that AI companionship causes poorer mental health. It may be that people who are already lonely or struggling are more likely to turn to AI in the first place. But it does highlight an important limitation: an AI system may provide a sense of presence without being able to offer the reciprocity, responsibility and unpredictability of a real human relationship.⁷
There may be times when that sense of availability is genuinely comforting. But if AI begins replacing difficult conversations, relationships, therapy or other forms of human support, the question changes. It is no longer simply “is this helping me right now?” but “what might I be moving away from by relying on it?”
So, can AI have a place between therapy sessions?
Potentially, yes , but probably as a supplement rather than a substitute. A 2025 systematic review and meta-analysis of 14 randomised controlled trials involving 6,314 participants found a small-to-moderate average improvement in mental-health outcomes among people using generative-AI mental-health chatbots. However, results varied considerably between studies and populations, so the findings should not be taken to mean that all AI mental-health tools will produce the same benefits.³
Used thoughtfully, AI may help someone organise their thoughts before a session, reflect on what has come up afterwards, journal, generate questions to explore, or notice patterns they want to bring back into therapy. But that is very different from asking an AI system to diagnose, make major decisions, assess risk, or replace professional support. The MHRA advises people using digital mental-health tools to ask what a product actually claims to do, whether there is evidence behind it, who it is designed for, and what happens to the personal data they enter.
Perhaps the most useful way to think about AI between sessions is not as another therapist, but as a reflective tool. Something that may help you find words, ask questions or hold onto a thought until your next session , while recognising that the meaning of that reflection still belongs to you.
Used in that way, AI does not have to compete with the therapeutic relationship. It may simply become one of the many tools a person chooses to use around it.
AI is unlikely to disappear from conversations about mental health, and perhaps the most useful response is neither to fear it nor to hand ourselves over to it. Like any tool, its value may depend on how, why and when we use it.
There may be something genuinely useful about having a space to organise a thought at midnight, find language for a feeling, or make a note of something you want to explore when you next meet your therapist. But there are also parts of therapy that cannot simply be generated on demand: being known by another person, experiencing disagreement or uncertainty, noticing what happens between two people, and having someone who is ethically and professionally accountable for the work.
Perhaps AI is most helpful when it turns us back towards ourselves rather than asking us to surrender our judgement to it. If it helps you notice something, question something, or bring something meaningful into your next therapy session, that may be useful. If it begins replacing human connection, professional support or your own capacity to make sense of your experience, that deserves a little more curiosity
When you reach for AI for emotional support, what are you hoping it will give you: information, reassurance, reflection, or connection?
You may also find: What Actually Happens in Your First Counselling Session?
References and further reading
British Association for Counselling and Psychotherapy (2026). Young people turning to AI as loneliness takes toll on mental health.
2. Brown, S. (2025). The big issue: What every therapist needs to know about AI. Therapy Today, BACP.
3. Zhang, Q., Zhang, R., Xiong, Y., Sui, Y., Tong, C. & Lin, F-H. (2025). Generative AI Mental Health Chatbots as Therapeutic Tools: Systematic Review and Meta-Analysis of Their Role in Reducing Mental Health Issues. Journal of Medical Internet Research, 27, e78238.
4. British Association for Counselling and Psychotherapy (2026). Ethical Framework for the Counselling Professions 2026.
5. Rahsepar Meadi, M., Sillekens, T., Metselaar, S., van Balkom, A., Bernstein, J. & Batelaan, N. (2025). Exploring the Ethical Challenges of Conversational AI in Mental Health Care: Scoping Review. JMIR Mental Health, 12, e60432.
6. Medicines and Healthcare products Regulatory Agency (2026). MHRA issues new guidance for people using mental health apps and technologies. GOV.UK.
7. Zhang, Y., Zhao, D., Hancock, J. T., Kraut, R. & Yang, D. (2026). Interaction with AI companions and psychological well-being. Nature Human Behaviour.