Is it weird to talk to AI about your feelings?

No. It's one of the most common things people do with AI, and one of the least discussed. In Harvard Business Review's 2025 analysis of how people actually use generative AI, therapy and companionship came out as the single most common use, up from third the year before. Meanwhile, plenty of the people doing it open with "this is going to sound stupid, but." The behaviour isn't unusual. The silence around it is.

If it feels weird anyway, that feeling is worth a look, because it usually isn't about the technology.

Why it feels weird anyway

Notice the label gap. When someone uses AI for work, they call it a workflow, and there's a bit of pride in it. When someone uses it to think through their life, they call it embarrassing.

Same behaviour, opposite label. What's uncomfortable usually isn't the AI. It's admitting the thing underneath: you needed somewhere to put a thought, and there wasn't an obvious person to call. That's not a character flaw, it's just true of most adult lives. Your friends get the edited version, the group chat gets the jokes, and the unedited version still has to go somewhere.

Help you're a bit embarrassed about is still help.

What it's actually good at

Being honest in both directions.

What works. Listening without an agenda. Asking the second question instead of rushing to reassure you. Helping you untangle a loop you've been circling for weeks. Being available at the moment the thought arrives, which is rarely business hours.

What it isn't. Therapy, a diagnosis, or a substitute for the people in your life. A good AI conversation should hand you back more of your own judgment, not less: clearer on what you think, readier to have the real conversation with the real person. If a tool leaves you feeling more dependent and less capable, it's the wrong tool. More on that here

If you're going to do it, three things matter

Know where your words go. Not every tool treats what you type as yours. Work accounts can often be read by admins. Some free tools use conversations to improve future models. You don't need to become an expert, you just need to read the one page where the company explains it, and be suspicious if there isn't one. Here's ours

Memory changes the whole thing. Starting from zero every time keeps it shallow, because you spend the first half re-explaining. Something that holds your context, the family drama, the tough boss, the thing you said three weeks ago, is what lets the conversation go anywhere.

Pick a space that's meant for it. A general assistant is built to complete tasks for anyone. That's a different job, and you can feel the difference in what it asks you.

The part worth saying out loud

A lot of people hide using AI this way. Some clear the chat afterwards, some keep it in a browser tab they never sign into, some just never mention it.

That's what people do with anything that helps before it's socially acceptable, and it's not a sign there's something wrong with you. It's a sign the thing is new. You're allowed to keep what helps.

The short version

Talking to AI about your feelings isn't weird. It's common, it's useful for a specific kind of thinking, and it isn't therapy. The only parts worth being careful about are where your words go, whether the thing remembers you, and whether it actually helps.

Rhythm was built for exactly this. A personal AI for thinking things through, with a voice you set yourself across five settings and memory that carries between conversations. Stored so no human can read it, nothing sold or shared, deletable from a button in settings. You can start without an account.