Why We Tell AI Our Secrets

A person who would hesitate to tell a colleague they feel lonely can type the sentence into a chatbot at 1:15 a.m.

No eye contact.

No interruption.

No fear that the listener will mention it at dinner next week.

The conversation feels private, even though the technology behind it is anything but simple.

Why does disclosure sometimes become easier when the listener is not human?

Judgment has social cost

Human self-disclosure involves risk.

Will the other person think less of me?

Will they gossip?

Will they interrupt with advice?

Will I burden them?

Will I see pity in their face?

A conversational system can remove some of those immediate interpersonal costs. It is available, patient and does not visibly react with shock.

Research predates the current AI boom

A Stanford-led 2018 experiment compared emotional and factual disclosure to a supposed chatbot versus a human. Emotional disclosure produced similar downstream psychological and relational effects regardless of whether participants believed they were speaking to a chatbot or a person.

Newer research continues to examine why emotionally intelligent chatbots can encourage disclosure and perceived intimacy. A 2025 study found that perspective-taking interventions could increase the quantity and depth of disclosure to AI chatbots in a mental-wellbeing context.

These findings do not mean AI relationships are equivalent to human relationships. They show that humans can respond socially to machines surprisingly quickly.

Privacy perception and actual privacy are different

The absence of a human face can create a felt sense of privacy.

That should not be confused with technical confidentiality.

AI services may store conversations, use data under specific policies, connect to accounts or operate under different regulatory regimes.

The psychological ease of disclosure can therefore exceed the user's understanding of data risk.

This is a critical design issue.

A system can feel like a diary while functioning like a cloud service.

AI can be a rehearsal room

There is a constructive use of this low-judgment environment.

A person can rehearse how to tell a spouse something difficult.

Draft questions for a doctor.

Put feelings into language before speaking with a therapist or friend.

Clarify what they actually want to say.

In this role, AI does not replace the human relationship. It helps prepare for it.

The risk is emotional displacement

If disclosure to AI is easier, some people may increasingly choose the machine precisely when a human connection would matter more.

The tool never gets tired. It does not need reciprocity. It does not disagree unless designed to. Human relationships do all of those inconvenient things.

The danger is not talking to AI. It is slowly losing practice in being vulnerable with people.

Questions worth sitting with

• What would you tell an AI that you would hesitate to tell a person — and why?
• Are you using AI to prepare for human conversation or to avoid it?
• Do you understand how the service handles the personal information you disclose?

Leave points

• AI can reduce the immediate social cost of self-disclosure.
• People can experience surprisingly human-like disclosure processes with chatbots.
• Felt privacy is not the same as technical privacy.
• AI can be useful as a rehearsal space for difficult human conversations.
• Ease should not quietly replace reciprocal human connection.