You already know this version of yourself. The one who responds quickly, explains things clearly, holds the thread of a meeting together. The one who doesn’t visibly fall apart. The one who, when someone asks how things are going, gives a considered answer that leaves the other person feeling reassured rather than burdened.
In 2020, a study published in Psychology Research and Behavior Management examined what researchers called the cost of impression management: the sustained effort of presenting a favorable version of oneself to others. One finding stayed with me. “A good impression manager does not necessarily have a good life.” The mechanism, in part, was relational. When you consistently signal high competence, the people around you take the signal at face value. “Their friends and colleagues are less likely to give them support,” the study observed, “as they observe little need.”
Now read that alongside what is happening in chat windows. A growing body of research on AI disclosure points to the same pattern from the other direction: people are telling chatbots the things they will not say to anyone who knows them. The two findings braid together into something specific. The audience that needs you to be fine is also the audience you cannot be honest with. And so, at some point, you opened a chat window and typed something you hadn’t said to anyone.
Not something small. Not a complaint about the day. Something from the category you usually don’t access in company, because the company is either counting on you, or judging you, or both. And it came out easier than you expected. Because the audience, this time, had no stake in the image you’d built.
In tech contexts especially, the performance is relentless. The expectation is not just competence but visible confidence: a facility with ambiguity, the ability to hold complexity without it showing on your face. You learn to answer questions before you have fully processed them. You learn that hesitation reads as uncertainty and uncertainty reads as weakness. You learn the shape of a sentence that sounds certain even when you are not. It becomes automatic. In a meeting, on a call, in a message that has to land well, the capable version of you assembles itself before you consciously decide to put it on.
This has a cost that tends to go unnamed.
There is the trap. Not that people don’t care. It is that a maintained image becomes its own barrier. You built it to succeed, or to protect yourself, or because the context required it. And then it started working too well, and the people closest to you stopped being able to see through it. Not because they weren’t paying attention. Because you are very good at this.
The explanations for why people disclose more openly to AI than to the people around them vary, but they converge on one point: the absence of social consequence. AI doesn’t carry the disclosure forward into the network of relationships that surrounds you. There is no version of tomorrow in which what you said becomes part of how someone who matters to you understands you. No colleague who files it away. No partner who references it three weeks later in a different argument. No reaction, however carefully managed, that quietly reorganizes the space between you.
For people who have been performing a particular self for long enough, that absence is not incidental. It is the specific thing that makes it possible to say something true. The performance requires an audience invested in it. Remove the audience, and the performance is no longer obligatory.
What this means in practice is something like: you open the chat because you are tired. Not necessarily of the work itself. Just of the version of yourself the work requires. The person who has already processed the complications. Who others in the room are counting on to have the answer before the question is finished. You type the thing not because the AI will do anything useful with it, but because you are exhausted from being the one who has it handled, and for a moment you need somewhere to put that down.
I know this territory from two directions. As a researcher who studies how people form emotional relationships with roles and identities, and as someone who has sat at a desk at the end of a long day and typed something into a chat window that I had not been able to say to anyone who knew me. Not because they would have been unkind. Because saying it to them would have required staying in the moment of being not-okay long enough for them to respond to it. That response, even if warm, would have meant something. It would have changed the shape of the relationship, however slightly. With AI, there is no such weight. I could say the thing and close the window. The performance could resume.
The relief is real. The reason for it is the problem. If the easiest place to be honest is a context with no stakes and no memory, then the contexts with stakes — the partner, the friend, the colleague who would actually be in your life tomorrow — have become places where honesty is too expensive to attempt. That is not a story about AI. It is a story about what professional life has done to the rest of life. The chat window is not closing the gap between your performed self and your actual state. It is making the gap survivable, which is a different thing, and arguably worse, because survivable gaps don’t get fixed.
So notice what the relief is telling you. It is not that you have found a better listener. It is that the people around you have been quietly conscripted into an audience, and you have been performing for them so long that the only place left to be a person is somewhere no person is.
If what’s described here reflects something heavier than tiredness, speaking to a professional therapist is worth more than a chat window.
Produced with AI assistance. Reviewed by the Silicon Canals editorial team before publication. See our about page.
About this article
This article is for general information and reflection. It is not professional advice. For your specific situation, consult a qualified professional. Editorial policy →
