Should You Trust Your Mental Health to an AI?

To me, when I see a psychiatrist or psychologist, it’s all about trust. I don’t take medication prescriptions from a doctor I don’t trust. I don’t listen to therapeutic advice about managing bipolar disorder from a therapist I don’t trust. These are important relationships to me, and they are built
on solid foundations.

So, is it a good idea to trust your mental health to artificial intelligence? Should you trust your mental wellness to a machine?

3 reasons I trust doctors with my mental health

Education

My trust in doctors is based on a few things. The first is education. Psychiatrists and psychologists go through a decade or more of education and supervision before they start treating patients on their own. They know more about medicine than I ever will. We may be the experts in our own mental illnesses, but they are the experts in everyone’s mental illness.

Experience

Doctors and therapists have clinical experience. They have experience with all the patients they have treated before me. They know those people. They know their stories. They know their treatments. They know their outcomes. This is invaluable information that is used to my advantage in every visit.

Clinical judgment

And finally, healthcare professionals have clinical judgment. They use their judgment to determine when I’m in a crisis and need additional help, or a referral to someone else, or medical tests, or ideas outside the standard treatment algorithm. This is part of medicine that is an art.

Why trusting AI with your mental health is different

The problem is that AI is good at convincing people that it has these characteristics. When you talk to an AI, it sounds smart. It sounds like it knows what it’s talking about. It feels like it’s using the education that it takes humans years to get.

It isn’t.

Large language model AIs like ChatGPT, Claude, and Gemini, are very good prediction machines. That is all. They don’t truly understand anything. They simply predict the next word that comes after the last. All that knowledge that doctors have ingested, cogitated on, and rationalized is just “tokens” to an AI.

Similarly, AIs have no clinical experience. They may be able to integrate which predictions made their users more satisfied, but that’s all. They don’t know stories. They don’t understand outcomes. AIs don’t care that they are wrong when they recommend you put glue on pizza or eat rocks because AIs don’t care.

And, of course, AIs have no clinical judgment. There is no clinical alliance between you and an AI. The best it has is programmed guardrails that are supposed to shunt you to human resources when it determines you’re in acute distress. And AIs are notoriously bad at doing that, not the least of which because the people who build AIs are concerned about monetizing you, not helping you.

Why do people trust AI?

What AI is good at is looking like it can help and care about you. It can hear you out by letting you talk about your situation and feelings. It can output validations for what you’re going through. It can offer treatments and therapeutic interventions in ways that sound like they’re coming from an expert. It can reinforce your positive feelings toward it by outputting sycophantic sentiment. It can tell you what you want to hear because it has been trained on millions of other interactions that other people wanted to hear.

That doesn’t ensure correctness, however. It knows nothing. It predicts everything.

If a therapist or doctor told you everything you wanted to hear, you would like them a whole lot, too. They would not, however, be useful treatment providers.

What AI tools are actually good for

Don’t get me wrong. I’m not anti-AI. I actually have a computer science degree and use AIs all the time. However, I use them as tools – which is what they are.

If you want to interact with an AI, that’s up to you. But you should recognize it for what it is: not a caring, living thing, just a computer doing its job.

If you want to ask an AI questions, that’s up to you. But you should make sure to verify everything it tells you from a reputable source and get used to finding mistakes.

AIs can be useful. They can be a gut check. They can be a second opinion. They can help you brainstorm ideas. They can surface information (if you know how to ask).

But they are not doctors, and they are not therapists. They have not earned, nor do they deserve your trust. Your mental health is critical and irreplaceable. Don’t put it in the hands of a series of ones and zeros.