Telling ChatGPT what you tell no one
Some people have a window on their laptop where they write things no friend has heard. A fear about a marriage. A thought about a parent. Something from years ago. Then they wonder whether it is strange that a chat assistant knows more about them than anyone they live with.
Most answers to that question are lists of what you should never type into a chatbot. This page asks something else: why is it easier this way, and what does saying it give you.
A psychologist who studied telling
Sidney Jourard was an American humanist psychologist. In the late 1950s he began measuring what people reveal about themselves, and to whom. In 1964 he collected the work in a book called The Transparent Self.
His argument was unusual for its time. A person no one knows pays for it with their health. And someone who is never known by others slowly stops knowing themselves.
He did not think hiding was a small thing. One line in the book puts it bluntly:
"I have known people who would rather die than become known."
— Sidney M. Jourard, The Transparent Self (Van Nostrand, 1964), p. 22
Why the machine is easier
A person listening has a face. The face can flinch. A friend may repeat the story, or answer it with a story of their own, or remember it at the wrong moment years later.
A chat window has none of that. It does not look shocked, and it will not bring the subject up at dinner. For many people this is the first place where the sentence gets said in full.
In Jourard's terms this is already disclosure, even though no person has heard it. Something that lived only as a feeling now has words, and the words can be read back.
What happens to what you said
Once it is said, three plain questions remain. None of them is about whether ChatGPT is good or bad.
Who you were talking to. ChatGPT is a general assistant. The same window writes emails, fixes code and plans trips. A conversation built for therapy works by other rules: it asks more than it answers and keeps its attention on you, not on the task.
Who holds the key. ChatGPT conversations are stored in your account on OpenAI's servers. OpenAI can technically read them. In an app where the archive is encrypted on your device, only your password opens it.
Whether the words come back. A disclosure that nobody returns to stays a confession. A general assistant remembers details so it can help with the next task. Therapy goes back to what was said: the same fear, heard again three weeks later in different words.
Two other pages cover nearby ground: how a general AI chatbot behaves as a therapist, and whether it is safe to talk to AI about your problems.
The question that stays
Whether it is weird matters less than it seems. What you told the window is now in words. Jourard would ask what you learned about yourself when you finally wrote it down.
What have I said to a machine that I have not said to anyone, and why did it have to be a machine?
Epimelea is an app for therapy with an AI therapist. It is a language model that works in the existential school. The conversation carries over from one meeting to the next, and only your password opens the archive.
About the app: an AI therapist for long-term therapy
This page is about self-disclosure and how conversations with AI are stored. It is not medical advice and not emergency help. If you are in danger right now, contact your local emergency number, or find a helpline in your country at findahelpline.com.