Nobody talks about why so many people find it easier to be vulnerable with a chatbot than with their closest friends — but the reason makes complete sense once you hear it

At some point in the last few years, a significant number of people have found themselves saying something to an AI chatbot that they had not said to anyone close to them. Not because the chatbot was a better company. Not because it offered superior counsel. But because, in the particular moment of typing it out, the chatbot felt like the only place it could go without cost.

The instinct is to frame this as strange, possibly worrying, a sign of something gone sideways in modern social life. That instinct is understandable. It is also a misreading of what is actually happening. The reason people find it easier to be vulnerable with a chatbot is not mysterious, and it is not pathological. It is a logical response to a set of social conditions that most people carry into every intimate conversation they have, and with a chatbot, for a moment, do not.

The problem has a long history

In 1966, Joseph Weizenbaum published a paper in Communications of the ACM describing a program he had created called ELIZA. The program was a simple text-based system that reflected statements back to users as questions, simulating, in a rudimentary way, the responses of a Rogerian therapist. Weizenbaum expected users to see through it immediately. Instead, he watched people engage with ELIZA as if it were real. His own secretary, he later described, asked him to leave the room so she could speak with the program in private.

Weizenbaum was disturbed by this. His 1976 book Computer Power and Human Reason (W.H. Freeman) was in part an extended reflection on what the ELIZA responses revealed about human longing: not that the machine was good at conversation, but that people had things they needed to say and were looking for somewhere to say them that felt safe. ELIZA was not safe in any meaningful sense. It simply removed certain risks, and that turned out to be enough.

Researchers Clifford Nass, Jonathan Steuer, and Ellen Tauber, in a 1994 paper presented at the CHI conference, documented what they called the Computers Are Social Actors finding: that people apply social rules and expectations to computers, even when they know they are talking to a machine. They speak politely. They respond to apparent moods. They feel, in some register, that they are in the presence of something like a social entity. This does not require delusion. It appears to be a default feature of how human social cognition operates.

The modern chatbots are not ELIZA. They are far more capable, more responsive, better at the surface mechanics of conversation. But the dynamic Weizenbaum identified in 1966 is still the underlying one, and understanding it explains the vulnerability question directly.

What the chatbot removes

Human vulnerability is not simply difficult because sharing something private feels uncomfortable. It is difficult because of what sharing with a real person actually involves. When you tell someone who knows you something that matters, you are not just transmitting information. You are changing the image of yourself in the mind of a person whose image of you affects your daily life. You are creating a new category of shared knowledge that could be recalled at any point, repeated, or used as a frame for how they understand you going forward. You are accepting the possibility that what you share will alter the relationship in a direction you cannot fully predict or control.

These are not paranoid concerns. They are accurate accounts of how human relationships actually work. The risk of disclosure is real, and the magnitude of that risk scales with how much the relationship matters to you. The closer the person, in many cases, the higher the stake of being seen differently by them.

A chatbot suspends all of this. It will not change how it sees you, because it does not have a persistent picture of you that carries weight in your life. It will not tell anyone. It is not carrying the information into future interactions as a changed understanding of who you are. It cannot be burdened, or worried for you, or pull away, or process what you’ve said in a way that reshapes the dynamic between you. The asymmetry that makes vulnerability with close friends so difficult, the fact that their reaction to what you say genuinely matters, is absent. And in that absence, there is space.

What you find in the space

What the chatbot actually provides, in these moments, is not connection. It is a temporary suspension of the evaluative pressure that surrounds ordinary disclosure. In that suspension, something often becomes clear: that the thought or feeling has been waiting for somewhere to go, and has been waiting because everywhere it might have gone carried social cost.

This is useful information. Not about the chatbot. About the person using it, and about the relationships around them. The fact that something felt sayable to a chatbot and not to the people closest to you says something about the specific conditions under which human disclosure feels possible. The chatbot did not create a relationship. It revealed, by its absence of relational stakes, what the relational stakes in real relationships were preventing.

People who use AI systems as a first step, before bringing something to a real conversation, are often doing something coherent: rehearsing, clarifying, building enough familiarity with the shape of what they need to say that the human version of the conversation becomes possible. That is a reasonable use of a reasonable tool.

What the chatbot cannot offer

The chatbot cannot know you in the sense that matters. It can process what you tell it, but it does not carry you in its mind between conversations, does not notice when you are quieter than usual, does not bring to the exchange the years of accumulated understanding that make a close friendship different from a helpful stranger. The reason its responses sometimes feel so adequate is that they are organized entirely around what you have just said, without the complexity that an actual relationship brings. That is not a substitute for what the relationship brings. It is the absence of it, experienced temporarily as relief.

Vulnerability with the people who matter to you remains the thing that builds them into the people who know you. The chatbot can hold a thought while you figure out what to do with it. The figuring out, and the doing, still happens with people.

Weizenbaum’s unsettling discovery was not that people confused ELIZA for a therapist. It was that people had things they needed to say and had not yet found a place to say them. Sixty years later, the landscape has changed considerably. The same cannot be said for the need that made his secretary ask him to leave the room.

If what draws you to those conversations feels like more than convenience, it may be worth exploring with a therapist what you’re looking for that hasn’t yet found a place to go.