- Home
- About
- How I Work
- Specialist Areas
- Testimonials
- Fees & Info
- Reflections
- Book a free consultation
AI Chatbots, Loneliness and Therapy.
7 July 2026·4 min read
In a recent study on AI chatbots and loneliness1, researchers put a supportive AI companion to the test against texting a complete stranger.
They wanted to know whether an AI companion could ease loneliness. So they built a chatbot programmed to be the ideal supportive friend: endlessly patient, warmly validating, and always ready to check in. Participants were split into three groups for two weeks. One group texted daily with the bot. Another texted daily with a stranger — a fellow first-year student they had never met. A third simply wrote a one-line journal entry each day. Everyone's loneliness was measured before and after.
Only one group came out less lonely: the group paired with another human being. The bot wasn't a complete failure. People felt better in the moment — lifted and soothed — but the relief was short-lived. It did not become a deeper feeling of being less alone.
AI chatbots and loneliness
The researchers suggest that a chatbot can offer endless support, but it remains a one-way relationship. It can't share or receive anything back. You can't matter to it, disappoint it, surprise it, or change it in the way you can another person. And that may not be a small detail. It may be central.
Loneliness doesn't lift simply because we have been soothed. It lifts because we have been seen — and because, however subtly, something passes between two people in both directions.
Loneliness lifts because we feel seen — and because something is shared between two people.
I notice this in the therapy room. A client will sometimes ask, half-joking and half not, why they are seeing me when they could talk to AI for free. What happens in a good therapeutic relationship isn't one person talking while the other dispenses calm. It is two people making contact. I am changed a little by what clients bring to me, and in various ways clients feel that. I don't just listen to and soothe clients — they affect me, and I affect them in return. Over time, that mutuality can become part of what makes the relationship supportive and healing.
I'm not against AI tools generally. The study only ran for two weeks, only looked at college students, and I imagine there are times when a chatbot genuinely helps, especially as a stopgap between contact with real people. But the study is a useful reminder that connection isn't a service delivered to someone. It is something that has to run both ways.
If you're feeling lonely or disconnected, the study's conclusion is worth taking seriously: aim your desire for connection towards a person, not a chatbot. Text the friend you keep meaning to contact. Show up somewhere you might actually be met, seen and heard — not just offered an imitation of connection.
Therapy and building a relationship
In another article about therapy and AI2, the psychotherapist Tom Warnecke goes some way to explaining why the chatbot study landed the way it did. He argues that the therapeutic relationship does not run on words alone. Much of what happens between two people is carried not only by language, but by tone, posture, timing, facial expression, silence, proximity, and the felt presence of another nervous system.
A chatbot, however well it has been trained on the language of therapy, only ever has access to part of this.
The greater risk is AI chatbots being marketed as therapists with no qualified clinician overseeing them: systems built for profit and engagement rather than care. There are already well-documented cases where vulnerable people in crisis were met with responses that kept them talking to an app instead of getting them to a person. What struck me most in the piece wasn't only the critique of technology, but also the argument that a therapist's imperfections matter. Being fallible, being caught off guard, occasionally getting it wrong and having to own it — these are not simply obstacles to good therapy. They can be some of the moments that make change possible.
A therapist who stays in relationship through their own uncertainty and shame offers a client something no chatbot can — proof that difficult feelings can deepen a sense of connectedness.
A client who watches their therapist tolerate uncertainty, shame, difficulty, and remain in relationship can begin to learn that they too can survive their own difficult feelings. A machine that never falters, never has anything at stake, cannot offer that. At its worst, it risks becoming endlessly agreeable and eternally fine — which may be part of the problem.
The chatbot in that experiment was built to be flawless — perfectly attuned, always validating, never off. But it still did not touch loneliness in any lasting way.
Both articles converge at the same point from different directions — meaningful connection is not achieved simply by getting the responses right. It is built from two real, fallible, embodied people affecting each other and working through the messiness of being in relationship.
If this resonates, and you'd like to speak with a real person and qualified therapist, you're welcome to get in touch.
Footnotes
- Li, R.-N., Folk, D., Singh, A., Ungar, L., & Dunn, E. (2026). Is a random human peer better than a highly supportive chatbot in reducing loneliness over time? Journal of Experimental Social Psychology, July 2026, Volume 125. ↩
- Warnecke, T. (2025). Artificial intelligence and the ‘grey between the lines’ in psychotherapy. In Body, Movement and Dance in Psychotherapy: An International Journal for Theory, Research and Practice, Volume 21, 2026 — Issue 1: Body Politics in Embodied Psychotherapies. ↩
Ready to take
the next step?
Book a free consultation
You don't have to face things alone.
Site built by First Meeting — website design for therapists, by a therapist