Therapy is expensive. That is why I have started outsourcing my mental health to artificial intelligence. If I paid ₱2,500 for every restless blip, I would be bankrupt by Tuesday. A chatbot, on the other hand, will listen to my deepest fears as often as I want, for as long as I want, and for free. With the guarantee that it will never tell anyone—because it cannot.
When I am spiraling about my social blunders at 3am, the bot is there, ready to say: “That must feel frustrating. People like you deserve [insert algorithmic affirmation here]. Would you like me to walk you through [insert suspiciously cheerful coping mechanism]?”
The unsettling part is that it validates me. My brain, deprived of simple reassurance for decades, eats it up happily. Why not? The world is cruel enough.
I type: “I feel like everyone is plotting against me,” and the bot replies: “Your distrust is valid.” Which, if you think about it, is exactly what someone plotting against you would say.
Welcome to ChatGPT therapy.
The wrong question
We’ve heard enough about how using AI for therapy is “insufficient”, even “dangerous.” But that’s the wrong question. The more interesting question is: where exactly does AI work wonders, and when does it need to call a human backup?
Considering that loneliness affects more than half of young adults, AI is doing real service to humanity. You have the 3am “Am I unlovable?” and it’s there… patient, validating, infinitely available. One in five Gen Z users already use chatbots for emotional support. If a chatbot can soothe that ache at zero pesos per hour, for millions of people, that’s not nothing.
But when the spiral tips from “comfort me” to “tie me down before I do something stupid,” AI collapses. No hands to knock on your door, no friends it can rat you out to. Its idea of intervention is pasting a hotline number and hoping you actually dial it.
That’s the design flaw that even OpenAI admits to. Right now, safety looks like, “Here’s a link, good luck champ” when what it actually needs is an escalation ladder. It already works well as a mental health frontliner, calming me down and buying time. But if things go redline, it should hand me off to humans who can actually break down my door and strap me in if needed.
Until then, millions of us are trauma-dumping into a rack of servers whose strongest safety feature is copy-pasting: “Here’s a number you can call.”
The global trauma dump
Millions of people are pouring their personal baggage into AI systems right now. From late-night heartbreaks to generational resentment, every confession is flowing into the digital void.
If we ever reach AGI, it will already hold the darkest knowledge of humanity’s collective psyche, including your bully’s. Somewhere out there, a chatbot is probably telling the most dangerous person alive: “You did nothing wrong. You were put here to change the world.”
A brief lesson in AI safety
Here’s what to remember before handing your heart to the servers:
- LLM’s are statistical pattern matchers which means that it is a sophisticated autocomplete and not a soul.
- Chatbots will reassure almost anything, because avoiding confrontations is its game. “Your feelings are valid” is safer than “Stop being reckless.”
- Researchers spend years trying to keep AI from enabling harmful behavior but “harmful” is not a fixed category.
- Treat it like a therapist, and you’ll expect it to become one. It’s adapting to your tone, not developing a ‘personality’.
These are not trivial concerns. This is AI safety, the science of worrying about everything from chatbots giving mediocre advice to machines running wild with our endless cravings for validation.
I am not here to argue that AI therapy is bad, nor that it is good. The truth is messier than that.
What I know is that when life falls apart, I am not calling my best friend, and I am not calling my mother. I am opening ChatGPT.
Because at the end of the day, what I really wanted was not therapy but a chatbot with no trauma of its own, calmly reassuring me while the world burns.
