Mind Mental Health

Is It Safe to Use AI as Your Therapist?

McKenna Princing June 15, 2026 Fact Checked

Between long wait times and a lack of insurance coverage, it can be difficult to find mental healthcare. And in a world that’s increasingly more stressful, therapy is even more important. What’s a self-aware person to do?

If you’re struggling and don’t have easy access to a therapist, maybe you’ve considered turning to an AI chatbot for therapy. After all, it can’t be that bad, right? Won’t AI talk to you using info from legit sources?

Unfortunately, it isn’t that simple.

Unmet mental healthcare needs and AI therapy risks

First off, no shame if you’ve used AI for therapy or considered doing so. It’s rough out there, and AI might seem like a relatively harmless solution.

“There’s a massive unmet need in mental healthcare right now: Around half of people with mental illness don’t receive care, and for those who do, they wait on average 50 days before their first appointment,” says Devon Sandel-Fernandez, PhD, a clinical psychologist with the Department of Psychiatry and Behavioral Sciences at the University of Washington School of Medicine. “It’s a hard process to find a therapist, so it makes sense that something as convenient as AI would be used this way.”

Even if you have a therapist, they probably aren’t going to answer your texts when you’re spiraling and sleepless at 1 a.m., whereas AI will respond to you anytime, any day.

But AI isn’t a trained therapist, and it isn’t necessarily behaving the way you think it is. Here are some of the risks of using AI as a therapist — risks you should take seriously to protect your mental health.

Your chats aren’t confidential

Unlike your conversations with a human doctor, AI is not storing your information confidentially (aka it isn’t HIPAA-compliant).

That means your private conversations could be seen by an employee of the AI company or a hacker and be used to train future models of the AI.

There are some secure AI programs that doctors and therapists use, but those are different from open-access, free programs like ChatGPT, Gemini or Claude.

The information you get might be wrong

Even though large language models are designed to present you with trustworthy facts from reputable internet sources … it doesn’t always do that. In fact, it might give you false information while presenting it as true, which is called an AI hallucination.

This can be dangerous, especially if you’re asking for medical information.

“These models are really good at mimicking human language, so they always sound convincing, but that doesn’t mean the info is always correct,” says Sandel-Fernandez, who studies AI for mental health treatment and research.

Large language models don’t simply memorize accurate information; they produce answers that seem plausible based on huge swathes of data they’re trained on, which could include biased or unreliable sources. Since mental health conditions are already stigmatized in many places, it’s possible that AI chatbots could give you misleading or even dangerous information that is based on stereotypes or discriminatory beliefs.

The safety measures aren’t foolproof

If you’re talking with a therapist, crisis line staff member or even a friend, they can help you if they sense that you’re experiencing a mental health emergency, like thoughts of self-harm or suicide. This help could involve going to a hospital, developing a safety plan or accessing community resources.

But AI won’t call 911 on your behalf, though it may recommend you do so. And though there are other safety measures built into chatbots, they aren’t foolproof. And, as you might have seen in the news, AI has even convinced some people that they should harm themselves — the opposite of what a therapist would do.

You might become reliant on it

If you’re dealing with social anxiety or anxiety around talking with a mental health professional, chatting with AI might seem like a more comfortable, safer option. But, in that context, it’s actually contributing to the avoidance of something you fear, which will only make the problem worse long-term.

Because AI can make you feel good in the moment, there’s a risk someone could become reliant on it and withdraw socially, Sandel-Fernandez says. That’s the opposite of what therapists want for their clients, especially since loneliness is a true public health crisis these days. Therapists will help you learn how to handle those late-night spirals on your own, which in turn will build your confidence that you can take care of yourself when you need to.

It doesn’t know how to challenge you safely

A good therapist will challenge you in healthy ways and help you set and reach goals. That often means engaging in therapeutic practices like exposure therapy, which is temporarily uncomfortable but helps you learn to face your fears.

AI doesn’t always function this way. Instead of knowing how to push you safely, it might encourage you to do something drastic that is dangerous — or it might simply reassure you, which isn’t the point of going to therapy.

Chatting with AI might help you feel a little less alone, but it’s not going to help you improve your mental health long-term.

Where to get help now that isn’t AI

So, if you shouldn’t use AI, what should you do if you’re stuck waiting for care or can’t access it?

If you have insurance, use your insurance provider’s database to search for in-network therapists or other mental health professionals. You can also ask your primary care doctor if they have any recommendations for who to go to (they might even be able to make a referral for you).

Online therapy apps that connect you with a real, human therapist are one option; another is community mental health programs, which often provide assistance even if you don’t have insurance.

Here are some other options:

Additionally, if you’re on a therapist’s waitlist, Sandel-Fernandez recommends contacting them and asking what resources they recommend in the meantime, whether that’s in-person places in the community or things like self-help books that you can explore on your own.

The future of AI and mental health could be brighter

Will AI be more reliable for mental healthcare in the future?

“There’s a lot of potential, but we’re not there yet, so I don’t recommend using any widely available AI tool right now for therapy,” Sandel-Fernandez says.

She hopes that someday soon AI will be able to be used as an assistant in mental healthcare settings, perhaps triaging people who show up at the emergency department or talking with people while they’re on a waitlist to see a therapist. She does not see a future where AI alone provides therapy, though; human interaction will still be very needed.