AI in therapy : Artificial intelligence is no longer something therapists can regard as a future issue. It is already here!
Like many other therapists, I am experiencing more and more clients attending counselling sessions referring to generative AI as being part of their support system and self-care - using it for things such as making sense of relationships, understanding their emotions, asking questions related to mental health and seeking support when another human being isn’t available.
And we now have UK evidence showing just how significantly AI is beginning to enter the therapy room.
The National Counselling and Psychotherapy Society (NCPS) Annual Member Survey 2026 describes technology and artificial intelligence as an increasingly important issue for the profession. More than half of respondents — 53.1% — believed AI was likely to affect their practice in some way. The Society concluded that the conversation appears to be moving from theory towards practice, while members continue to raise legitimate concerns about ethics, confidentiality, data security and protecting the human connection at the heart of therapy.
Another striking finding from the NCPS Annual Member Survey 2026 relates directly to clients' use of AI.
When counsellors and psychotherapists were asked whether clients had told them that they were using AI tools such as chatbots for emotional or mental-health support:
15.6% said frequently
43.1% said occasionally
18.1% said rarely
16.5% said never
6.7% were unsure
In other words, 58.7% of respondents said they were frequently or occasionally encountering clients using AI for emotional or mental-health support. If we include those who said they encountered it rarely, that rises to 76.8%.
And the NCPS findings suggest that the effects aren't entirely positive or entirely negative. Practitioners reported clients using AI in a constructive and positive way helping them with organising their thoughts, journalling and reflecting between sessions. It was also being used for psychoeducation and reassurance and could help clients find language to describe their experiences.
But there were also significant concerns about its use - self-diagnosis, over-reliance on AI explanations, misinformation being accepted as fact, expectations that therapy should provide quick answers, reduced engagement with the therapeutic relationship and AI reinforcing existing beliefs without sufficient challenge.
I feel that we are no longer talking about whether AI might eventually enter therapeutic work, for many therapy sessions, it already has. But the mixed picture referred to above is important because it suggests that simply categorising AI as either "good" or "bad" misses what is happening in practice.
It therefore has got me thinking – maybe the concerns and questions about this needs to move beyond:
"Should people be using AI for emotional support?"
And move more towards:
"Given that people are using it, how can therapists begin to embrace it’s use more and help clients to do so safely, critically and without losing what is uniquely valuable about human therapy?"
Perhaps therapists also need to begin considering how a client is using AI rather than whether it is being used in the first place.
As therapists, I feel that it is important to distinguish that AI is not therapy but that doesn’t mean it has no place in therapy settings.
A journal isn't therapy.
A self-help book isn't therapy.
A mindfulness app isn't therapy.
Talking to a supportive friend isn't therapy.
Yet all these things can potentially support someone's wellbeing.
So perhaps AI can sometimes be understood similarly: not as a replacement for therapy but as a tool to support therapy.
Let’s Stay Curious
When I am training my students at my counselling training centre, Northamptonshire Centre for Counselling Studies, I will often be found to be consistently encouraging my students to remain curious when working with clients. Perhaps this approach needs extending to a client’s use of AI
For example, imagine a client says:
"I've been talking to ChatGPT about this."
A therapist may respond with:
"You really shouldn't use AI for mental-health advice."
But perhaps a more curious response could be:
"That's interesting. What have you been asking it?"
This latter response is more likely to encourage a more helpful conversation about what they might be struggling with.
Another example, perhaps a client asks AI:
"Do you think I'm a bad person?"
Again, instead of stating that AI really shouldn’t be used for mental-health advice, I would suggest that it isn’t about whether the AI system responds with ‘yes’ or ‘no’. For me, it is more about remaining curious in relation to why the client needed to ask AI this in the first place.
The technology may be a new way of accessing perceived support by the client but the psychological processes underneath it often isn’t.
Is it About Safety?
AI is a relatively new entity in the grand scheme of the history of therapy. But perhaps we should consider that this it is now creating a space for clients who feel safer asking AI than talking to another human being directly.
Another reason to remain curious rather than dismiss the use of AI as not being a ‘real relationship’.
Instead, a curious response could be:
“What feels safer about telling AI this than telling another person?”
Here, we aren’t talking about using AI any longer but instead we are talking to the client about relationships.
Using AI between sessions
In my experience working with clients, I have also noticed that clients who have used AI between sessions have attended counselling sessions more focused, with organised thoughts and being able to verbalise better what is happening for them in the present moment. This was supported by the 2026 NCPS findings.
When isn’t using AI in therapy helpful?
I feel that it is important to provide a balanced view and acknowledge where using AI isn’t helpful, such as AI providing inaccurate information, oversimplifying things that may be a lot more complex and perhaps reinforcing potentially unhelpful beliefs.
The 2026 NCPS survey also found that therapists also reported clients developing expectations of quick fixes or definitive answers, comparing therapists' responses with AI, and potentially placing less value on the slower and more uncertain process of human therapy.
Should we feel the need to compete with AI as therapists?
Another finding in the 2026 NCPS survey was when members were asked about external factors affecting their practices, AI and digital alternatives were increasingly being mentioned as factors influencing client behaviour and potentially reducing demand for traditional private practice.
Therapists might experience this as threatening to their private practices or organisations they work for or manage. However, is competing with AI the answer?
Even though AI can provide an explanation of attachment theory in seconds, can be available at 4am or provide 20 journalling prompts instantly – therapists don’t need to compete with this. We provide something that AI can’t - a relationship. Remembering something significant from 6 months ago, noticing body language and silences, experiencing ruptures and repair, providing challenge and vulnerability – the client taking the risk of being genuinely know by another human being.
References
