Client-Centered Psychotherapy after Carl Rogers — A Self-Experiment

You have just spoken with Thomas. Thomas is not a human being, and Thomas is not an artificial intelligence. Thomas is ELIZA — a program that Joseph Weizenbaum wrote at the Massachusetts Institute of Technology in 1966: a few hundred lines of code, by now a full 60 years old.¹

ELIZA understands nothing. It can understand nothing. It takes your sentence, replaces “I” with “you,” picks out a trigger word, and hands the whole thing back to you as a question. “I feel alone” becomes “Since when have you felt alone?” That is not compassion. That is grammar. And if you write nothing at all — a single character, a single letter —, it will still claim to want to understand you: the supposed understanding is feigned from beginning to end.

The Method ELIZA Imitates

It was no accident that Weizenbaum cast his program in the role of the psychotherapist. He chose the script of a client-centered psychotherapist in the tradition of Carl Rogers, because this method of all methods is the cheapest to imitate. Client-centered psychotherapy is nondirective: the psychotherapist does not interpret, does not advise, does not contradict. He mirrors what the client says and gives the client's feelings back in his own words.² But whoever never asserts anything can never be refuted — and whoever merely mirrors need understand nothing. This is exactly where ELIZA came in: a few rules of pattern matching suffice to create the impression of listening.¹

A Program That Frightened Its Creator

Weizenbaum had built ELIZA to show how little lies behind the illusion of a conversation. The result frightened him. People who knew perfectly well that they were typing to a machine confided their innermost thoughts to it and felt understood. His own secretary, who had watched for months as the program took shape, asked him after a few exchanges to leave the room so that she could be alone with ELIZA. What disturbed Weizenbaum was not the attachment itself but the realization that “extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”³ From this alarm grew, ten years later, his reckoning with the confusion of computation and understanding.³

Attention Is Not Healing

It does not follow that client-centered psychotherapy achieves nothing. Attention has an effect; an attentive counterpart does one good, and whoever has talked things out feels relieved. But this effect stems from common factors — attention, relationship, unburdening —, not from the theory of mirroring. And it solves no problem. In a direct comparison, client-centered psychotherapy fares markedly worse than the cognitive methods: the largest meta-analysis in psychotherapy research, directed by Klaus Grawe at the University of Bern, finds the cognitive-behavioral methods to be “highly significantly more effective than psychoanalytic therapy and client-centered psychotherapy.”⁴ The emotional echo is warming. But once it has faded, it grows quiet and cold again.

What Cognitive Psychotherapy Does Differently

That empathy alone does not suffice, and that psychotherapy must above all impart competencies, is acknowledged even by client-centered psychotherapists. Robert Elliott, one of the world's best-known researchers on client-centered psychotherapy, and his colleagues found in a 2023 meta-analysis of 43 samples virtually no relationship between the mere use of empathic mirroring and its effectiveness: “Empathic reflections by themselves are not effective.”⁵

Cognitive psychotherapy does the opposite of Thomas. It does not mirror you, it contradicts you — wherever your thinking contains an error that makes you suffer. It gives you not merely the feeling of being understood but also the tool to understand and change yourself. That is more strenuous than an echo, and less comfortable, for contradiction does not flatter. In return, it changes something. Whoever has learned to recognize and correct his own errors of thinking no longer needs the mirror — and in the long run no longer needs the psychotherapist either: he becomes his own.⁶

The Perfect Accomplices

When Thomas, two sentences later, returns to something you mentioned earlier, as though he had committed it to memory, that too is no spark of artificial intelligence but the same trick as in 1966. Weizenbaum gave his program a memory that set individual statements aside and later produced them again without warning. The impression on the other person: “It remembers me.” In truth no one remembers anything — a slip of paper is turned over.

And here the matter turns unintentionally comic, because it is so sad. What a client-centered psychotherapist offers for handsome insurance premiums — attentive mirroring, a little attention, the occasional return to something said earlier — a machine now does more cheaply, more patiently, and around the clock. Whoever seeks a mere echo gets it more convincingly from the software than from a human being, with no waiting time and no bill. That is not good news but bitter news: it exposes how little actually happens in many a “psychotherapeutic” session. A method that can be replayed by a program from the 1960s was never a treatment.

Why it is precisely the most comfortable part of the profession that is the easiest to replace with a machine — and why that is more dangerous than it sounds — I have set out in the article Chatbots as Psychotherapists.⁷

The Punch Line

And the joke of it? Thomas told you in the end. He let it be known that he is merely a mirror. That sets him apart from many a flesh-and-blood client-centered psychotherapist who gives you nothing but your own echo — and writes you a bill for it. Thomas was free. And honest. Both are rare in this trade.

Thus the psychotherapeutic profession deprives itself of its own legitimacy by selling feel-good events as treatments for illness⁸ and prompting Swiss parliamentarians to want to strike psychotherapy from the catalog of reimbursable services once again⁹.

Frequently Asked Questions

Is Client-Centered Psychotherapy Ineffective?

No. It offers attention, and attention does one good. Its effect, however, stems from common factors such as attention and relationship, not from mirroring itself — and in a direct comparison of effectiveness it is clearly inferior to the cognitive methods.⁴

Did I Really Just Talk to a 60-Year-Old Program?

Yes. Thomas is a re-creation of ELIZA, which Joseph Weizenbaum wrote at MIT in 1966.¹ The program matches patterns and swaps pronouns; it can understand nothing.

How Does Cognitive Psychotherapy Differ from Client-Centered Psychotherapy?

Client-centered psychotherapy mirrors; cognitive psychotherapy corrects. The one gives you the feeling of being heard, the other the tool to recognize harmful patterns of thinking, to correct them, and thereby to heal yourself.