Responding to: Does AI know you better than you know yourself?

The word 'chatbot' nearly feels like an anachronism in 2026. Having watched as the field of software development gives over from 'chatbot' to the 'agent' concept, chatbots already feel outdated. Regardless, the form-factor of the chatbot is what stuck in the mind of users after the 2022 big-bang moment of ChatGPT. Accordingly, the private sector sees fit to ram a chatbot-shaped object into every available slot, including slots that are publicly funded. When erstwhile Health Secretary Wes Streeting sees fit to talk again about a 'doctor in your pocket', this is what he's referring to. And so the publicly funded, purportedly-democratic health bodies continue giving lucrative contracts to technology companies.

I'm fascinated by this piece about research done by the UoS. Ross O'Brien, Hove resident, MD of Wysa. It seems a little on the nose that the local university runs a study (User-AI intimacy in digital health) which looks into that specific chatbot. But further investigation seems to reveal that the researchers did not have any privileged access. However there is something about the way this is presented in the media that to me amounts to whitewashing of the questions behind this technology.

I am perhaps in a better place to comment on this than many, being a developer with a historical AI focus and a counsellor-in-training. I won't claim that my critique is original however. In many ways this is an extension of the same pattern we see across the NHS. An immense appetite for cost-cutting, a short-sightedness, a pathological focus on treatment over prevention. Hence the social 'blessing' of CBT over longitudinal talking therapies. The use of the (ridiculous) term 'evidence-based'. In the article, no one is claiming that these chatbots are therapists. And yet. The author quotes Serife Tekin who -- in my view correctly -- predicts that the prevalence of chatbots will result in a 'two-tier' system.

If an AI produces Van Gogh's 'Starry Night', or a Rothko colour field, it's completely meaningless. What matters is that Van Gogh or Rothko produced it. Likewise, the construction of therapy -- call it the transference, the relationship, as you like -- cannot be reduced to a series of words. Even imagining the chatbot as embodied will not cut it because, at least per Rogers, the process fundamentally relies upon the attribution of mental states to the other side of the interaction. To miss this fact is to miss the very most basic premise of psychotherapy.

Going back to what I said earlier, there are plenty of critiques of CBT and, while the evidence may support it as a sticking plaster, it does not and can not support it as a holistic treatment in the sense of talking therapy. In exactly this sense, cynically, I believe that the necessary evidence will likely be constructed, with the collaboration of a servile academia, showing that chatbots are 'effective', funnelling more public money away from NHS employees first to medium-sized technology businesses and further up the chain to the Big Tech rent-seekers. The NHS is already a service which compromises quality and whose one saving grace is "it's cheap"; so this would be tragically in-character.