What happens when talking to a chatbot feels safer than talking to a friend?
More people are using tools like ChatGPT to vent, process stress, and think out loud - not because the AI understands them, but because it doesn't judge, gossip, or get tired of listening. The reasons are consistent across people who do this: no social consequences, no reciprocal obligation, no risk that your words get used against you later.
This isn't a fringe behavior. Research on self-disclosure consistently finds that people share more personal information when they believe there won't be social consequences. A chatbot checks that box. It knows nothing about your life unless you tell it, has no mutual friends, and won't run into you at a party.
The Case For It
The argument isn't that AI provides genuine emotional support - it doesn't. Large language models (programs trained to predict what text should come next) have no inner experience, no persistent memory by default, and no actual opinions. When Claude says "that sounds really hard," it's pattern-matching against millions of human responses, not empathizing.
But that can still be useful. The act of articulating a problem in writing helps you process it - psychologists have studied this for decades under the label "expressive writing." The AI doesn't need to respond meaningfully for the writing itself to help you clarify your thinking. The chatbot is essentially a blank page that writes back.
For people with social anxiety, limited support networks, or topics that feel too sensitive to share with anyone who knows them, the option to vent to a machine can offer real, if limited, relief.
The Risks Worth Taking Seriously
First, privacy. What you tell a chatbot is logged, and in many cases used for model training. Treat it like a diary that might someday be read - not a sealed confessional.
Second, substitution. Using AI to avoid difficult conversations with real people is a short-term comfort that makes relationships worse over time. Venting to ChatGPT instead of telling your partner you're overwhelmed doesn't resolve the underlying problem.
Third, the validation loop. AI tools are very good at mirroring your perspective back to you, which feels validating. But that validation is hollow - the AI agrees with you because statistical agreement is likely in that context, not because your view is correct. It won't push back, challenge your assumptions, or notice when you're wrong. People who care about you will.
The habit to watch for is substitution: when AI conversations start replacing human ones rather than supplementing them. Using AI to think through a problem before a hard conversation is useful. Using it to avoid having the conversation at all is the pattern that causes damage.