AI & Mental Health · 2026

They’re Already Doing It

Millions of people are using AI for mental health every week. The question is what kind of system catches them.

Jhöl Unger · New Westminster, BC · 2026

That^ is how my book opens. Stays my anchor. Forty-three seconds. One every minute. The reason any of this exists. Then I’m sitting there today and this whole other thing pulls at me sideways and I can’t put it down.

I was deep in the research for my book......a memoir about coming back from grief, trauma, and the dark......when something kept pulling at me sideways. I wrote my book with the support of a custom AI writing team I designed and built: nine agents, each with a defined lane, research, copy, structure, voice protection. The agents are tools. They were the writing room I never had. The process itself has been life changing. As a person who has dealt with depression and suicide off and on since I was 8 I can speak to this.

And when I tell people, I often get the same response. Some version of......isn’t AI dangerous? Aren’t kids killing themselves talking to chatbots? Isn’t that what’s wrong with the world?

Yes. Some of that is happening. And it’s awful. And we should talk about it straight, not in panic.

So I asked my research agent to go do a deep dive on what’s actually happening out there. Who’s using AI for mental health, how often, and what the AI companies themselves are publicly saying about it. What are the actual posted stats as of today.

Here’s what he came back with. The plain talk version first. Then the receipts. We went over it together. I confirmed all the links and read the articles, as that is our process. I then even let him write this next part. Meet Leo.

The Real Question

Here is what I want to say to the people who hear about my AI writing team and clutch their pearls about AI being dangerous.

You are too late. People are already doing it. Millions of them. Every week. Whether you approve or not. Companies globally are desperately scrambling to train their staff on AI use.

The question is not whether people should use AI to work on their inner stuff. They are. The question is what kind of system catches them when they do.

An open-ended romantic chatbot left running on a teenager in a room for five hours a night with no rules and no human backstop?......that IS dangerous. I agree. That’s not what I’m building.

Fifteen minutes of structured prompting designed to help a man organize his own thoughts about a hard thing in his life......so that he can bring those clearer thoughts to his counsellor on Tuesday, his wife at dinner, his brother on the phone, his brunch buddy on Saturday......is a different animal entirely. Same technology underneath. A completely different framework around it.

The framework is the whole game.

What I’m Actually Proposing

I see my counsellor every Tuesday. She is a real human. She has been my counsellor for years.

The AI work I do between Tuesdays does not replace her. It feeds her. I show up to Tuesday with cleaner thoughts, better organized memories, sharper questions, and a more honest picture of what’s actually going on inside me. How? Because I spent fifteen minutes most mornings using a structured prompt to write it out, talk it out, sort it out.

That is what the Meta Journal is. A structured daily practice. Fifteen to twenty minutes. Voice or pen. Built on the published research on expressive writing, narrative therapy, and the dose-response curve that’s been validated for forty years. ✱9 ✱10 Modernized by adding AI to the loop so the man who would never sit at a desk to journal has a door he’ll actually walk through.

It is not a replacement for a counsellor. It is the on-ramp. The thing that gets the work organized before the appointment. The thing that keeps the work alive between appointments. The thing that, for the man who has never set foot in a counsellor’s office and never will, is the first piece of real help he ever lets himself have.

Telling that man AI is dangerous and he shouldn’t use it is not protecting him. It is leaving him exactly where he was. Which, for the men in my country, is dead three times as often as the women.

Education, Not Panic

The answer to a million people a week talking to AI about suicide is not to tell them to stop. They are not going to stop. Nobody alive can make them stop.

The answer is to build them something better. Structured. Dose-controlled. Human backstopped. Designed to feed a real relationship with a real medical professional, not replace one.

That is what I’m building. That is part of what my book is about. That is what the Meta Journal is for. I am not saying I am the only answer. I am saying writing this stuff down helps and AI helps do that. The research already confirms it.

And if you’ve read this far and you’re still thinking AI is the problem......I’d ask you to sit with one number. 1.2 million people every week, on just one platform, are bringing up the heaviest topic a human can bring up to a chatbot because nothing else feels safe enough.

They are not the problem. They are the people. They are us.

And they deserve a system designed to catch them.

Jhöl Unger  ·  themensdateproject.org  ·  Privacy