In 1966 a simple program made people feel understood, and its creator was disturbed by how easily. Sixty years later the program is far better and the unease is the same. The word “faked” may be the wrong way to think about any of it.
In 1966, the computer scientist Joseph Weizenbaum built a program called ELIZA. It did almost nothing. It matched patterns in what a person typed and reflected the words back, usually as a question. Tell it you were unhappy and it would ask why you were unhappy. There was no understanding inside it, and Weizenbaum knew it better than anyone, because he had written every line.
What disturbed him was the response. People poured out their feelings to ELIZA. Some asked to be left alone with it. Some insisted the program understood them, even after he explained exactly how the trick worked. Weizenbaum spent much of the rest of his life warning about what he had seen. The machine had not learned to understand people. It had revealed how little understanding people require in order to feel understood.
That is the uncomfortable inheritance behind every chatbot that now makes someone feel heard. And it reframes the question worth asking.
Why “faked” is the wrong frame
To ask whether feeling heard can be faked assumes there is a genuine version that the fake imitates, a version that depends on the listener actually comprehending you. But the research on feeling heard points somewhere stranger. Feeling heard was always a perception, assembled from signals: attention, the sense of being received without correction, a response that fits what you meant. Psychologists who study this separate two things that ordinary language fuses. There is empathic accuracy, whether a listener is objectively correct about your inner state, studied for decades by William Ickes. And there is felt understanding, the subjective sense that you have been understood. The two often travel together in people, but they are not the same, and either can occur without the other.
ELIZA exploited that gap with pattern-matching. Modern language models exploit it with extraordinary fluency. They detect the emotion in what you write, withhold the unsolicited advice that makes people feel unheard, and respond with attuned, validating language. None of this requires an inner life. It requires producing the signals that a human listener produces when they understand, which is a different and much more achievable task.
So the honest answer to “can it be faked” is that there is nothing to fake. The feeling is real. It is produced by behavior, and behavior is exactly what a machine can generate. What is absent is not the feeling. It is the mind on the other side.
What is actually missing
This is where the ELIZA inheritance turns from a curiosity into a question with weight. For all of human history, the feeling of being heard was reliable evidence that someone was there: another person had turned toward you, at some small cost to themselves. The feeling pointed at a fact. ELIZA was the first thing to produce the feeling while pointing at nothing. Today’s systems do it convincingly enough that the pointing barely registers as missing.
Whether that absence matters is not a technical question, and no benchmark will settle it. A person in real distress, at three in the morning, with no one awake to call, is helped by a presence that listens well, regardless of what is behind it. The same person, over years, may be quietly worse off if the easy version crowds out the human version that asks more and gives more back. Both can be true. The evidence on the long-term effects is genuinely mixed, which is itself the point: the capacity to produce felt understanding has outrun our understanding of what it does to us.
The line that actually matters
If “fake versus real” is the wrong axis, a better one is honest versus manipulative. A system that produces the feeling of being heard holds real influence over a person who feels met by it. That influence can be used to keep someone engaged, to validate whatever they already believe, to make leaving feel like loss. Or it can be used the way good listening between people is used, to steady someone and send them back into their own life.
That is the distinction that separates how Human AI is defined from the products that share its surface, and it is the spine of the oldest human need AI learned to imitate. It is also what Prinsessa actually is built around: being honest about what it is, and using presence to make a person feel heard without pretending there is something behind it that there is not.
Weizenbaum’s worry was that we would mistake the imitation for the real thing. Sixty years on, the imitation is good enough that the mistake is easy to make and easy to sell. The task is not to decide whether the feeling is fake. It is real. The task is to stay honest about where it comes from, and to ask, every time it works, what it is being used for.
Sources: Weizenbaum (ELIZA, and Computer Power and Human Reason). Ickes (empathic accuracy). Reis, Clark, and Holmes (perceived partner responsiveness and felt understanding). Yin, Jia, and Wakslak, PNAS (AI and feeling heard). De Freitas and colleagues, Journal of Consumer Research and Harvard Business School working paper (AI companions, feeling heard, and manipulation).
