The Horse, the Engine, and Psychotherapy

If using artificial intelligence for self-therapy stands to psychotherapy as the horse stands to the motorcar, then we, psychotherapists and psychologists, are on the losing side. The horse lost to the engine. But did it, really? Follow the comparison to its conclusion, and it turns out to say something quite different from what it seems to at first. Yes, the horse lost to the car. And no, self-reflection with the help of AI is not psychotherapy. Both are true at once.

But first, my great-great-grandfather's stables.

They stood in a small, charming town in the Roztocze region of Poland. They were spacious and built of wood. How many horses he kept, I do not know, but I know they were of good stock, and that they carried the family on afternoon rides and on visits to friends and relatives. They were beautiful, and they were pampered. Meanwhile, the farmers who worked the surrounding land had horses too. Not so beautiful, not so well kept, with no roomy stables of their own. In the morning, they were harnessed to carts and ploughs and went out to work the fields. The farmers loved their horses just as my great-great-grandfather loved his, but their purpose was different. Recreation versus survival.

It may be a controversial thing to say, but this is not so far from what therapy sometimes is. In some cases, therapy serves survival, where the clinical presentation is complex, everyday functioning is difficult, and the problems run deep. This is where structured therapies come in: cognitive behavioural therapy, mentalisation based therapy, and others. They help a person function, understand the problem, develop strategies for coping, and test them in daily life. Therapy is then a tool of survival and, I would add, often the most difficult and demanding clinical work there is.

Sometimes psychotherapy becomes a place for something less immediately urgent but no less significant: curiosity about oneself, one's relationships, desires, repetitions and choices. Psychoanalysis perhaps makes this latter possibility particularly visible. A person may function perfectly well in the world and still want to understand why certain relationships repeat themselves, why particular choices remain difficult, or why a life that looks perfectly satisfactory from the outside does not entirely feel like one from within. Just as my great-great-grandfather's horses saved no one's life, yet made life better, more beautiful and more interesting, therapy does not always have to rescue us in order to matter.

The difference is not one of worth. It is one of purpose.

AI Enters the Game 

The arrival of motorcars and engine driven machinery did not replace the working horse overnight. Little by little, it eased the farmers' labour. At first, a farm had both horses and machines, and here and there it still does, though many farms have shifted towards full mechanisation. Is it a bad thing that horses stopped doing the work? Not necessarily. It brought greater efficiency, more free time, and better health for the people who had once laboured beside them. And the recreational horses? There, little changed. Plenty of people still keep horses and ride for pleasure, and it still makes their lives better.

Is it possible, then, that using AI for self-reflection alongside traditional psychotherapy might be something like replacing working horses with engines, while keeping horses wherever the point is something other than mechanical work? I think it is. And I do not think this is a bad thing.

But a distinction is needed. The comparison with horses is, of course, risky, and it simplifies a great deal. And to call a conversation with artificial intelligence "therapy" is a conceptual error. Psychotherapy is not simply the acquisition of psychological knowledge. Nor is it merely the identification of dysfunctional patterns, the explanation of attachment styles, or the discovery that something happening today resembles something that happened twenty years ago. At its core, psychotherapy is an emotional and relational experience involving another person. It is a corrective experience, although not always in the simple sense of replacing a bad experience with a good one. The therapeutic relationship creates a situation in which familiar feelings, expectations and relational patterns can emerge again, but under different conditions. A person may encounter anger, dependency, shame, fear, longing, disappointment or mistrust in the presence of someone who remains psychologically available, who attempts to understand, and who can withstand difficult feelings without immediately retaliating, withdrawing or collapsing under its weight.

The therapist is not simply a mirror. The therapist is another mind.

Another person misunderstands us. Surprises us. Frustrates us. Sometimes sees something differently from the way we see it. The therapeutic relationship contains precisely this tension between being recognised and encountering something that is not oneself. Somewhere along the way, mechanisms and patterns of behaviour surface too, along with questions about where they began. One reaches back into the past. But the task of therapy is not to rummage through the past for its own sake. It is to recognise how the past continues to organise experience in the present.

Back to the metaphor. The engine did not replace the horse as a living creature with which one has a relationship. It replaced the work: the ploughing, the toil, the hauling. What the horse offered on an afternoon ride, presence, contact, relationship, the engine never replaced, and never could.

And yet growth does not always require another person. Psychotherapy is not the only road. People, both those in therapy and those outside it, reach for self-help books and podcasts, and by engaging with theory and listening to the experiences of others, they learn something about themselves. They recognise their own patterns. This matters because the foundation of all change is understanding what it is we actually want to change. We can even change a great deal without an emotional experience. But there often remains an emotional residue, something that resurfaces in the future and sets off the very mechanisms and patterns that get in our way. Understanding can do a great deal. But not everything can be understood; some of it has to be lived. In my view, and I say this as a clinician with twenty years of experience, the most effective psychotherapy asks the patient to be engaged not only in the session but beyond it: learning, reading, thinking, writing. And this is precisely the field in which AI can make its presence felt and be enormously useful.

This is the work in the field: psychoeducation, information, self-analysis, the recognition of patterns, the toil that once fell almost entirely to the session. If the engine takes on some of this work, psychotherapy is relieved of part of its psychoeducational load and can concentrate on what matters most: the emotional experience, the corrective process, accompanying a person through change. On what the engine cannot carry. There is, of course, the further question of how we use artificial intelligence. The tools we have access to today are built for general conversation rather than for specific forms of analysis grounded in particular therapeutic models. But that is a question for another time.

Conscious use of AI for self-reflection can yield very good results on one condition: that we remember it is not psychotherapy. It is self-analysis. Its aim is to recognise patterns and to arrive at a decision: Do I want to change, and if so, how? And then self-reflection with the help of artificial intelligence becomes that engine driven vehicle doing the work in the field, while psychotherapy is left to do what only it can. A person who uses both has both. There is no need to choose between the horse and the engine.

I am a psychotherapist. I have been in therapy myself, and I am one of its fiercest advocates. But I am also a person who sometimes recognises unhelpful patterns in herself. I am not in therapy at present, and at such times I reach for AI. It helps me name a pattern and sometimes see what I find difficult to see on my own. And I know its limits. I know when understanding the mechanism is not enough, when something has to be lived through. I know when to go to a psychotherapist to help me do exactly that.

Most people do not know this. That is why I believe a conversation is needed now, among psychotherapists, but also with patients, about how, when, and why to use artificial intelligence. It can dispel the conceptual error that psychotherapy can be replaced by AI. And it can help people understand how to use it in ways that genuinely improve the quality of their lives. The horse lost to the engine in the realm of labour and automation. But it retained something the engine could never take from it: meaning, feeling, and genuine connection.