Can AI Replace Your Therapist? Pros and Cons of AI Therapy

By William Schroeder, LPC, NCC and clinically reviewed by Teri Schroeder, LCSW

Last reviewed: October 2026

Can AI replace your therapist? No. ChatGPT and other chatbots can help you sort your thoughts, journal or practice a skill, but they aren’t therapists, and most of what you tell them isn’t private. The American Psychological Association said the same thing in a November 2025 health advisory: don’t use chatbots and wellness apps as a substitute for care from a qualified mental health professional (APA, 2025).

I’m a licensed counselor and co-founder of Just Mind, an independently owned group practice in Austin. Before I was a therapist I worked in IT, and I use AI every week. I also work on Texas mental health policy. So I don’t mean to badmouth the AI stuff. My big thing is making sure buyers beware.

Below is what I’m seeing, where I think AI helps, where it worries me, and what I’d tell a parent.

What Are Therapists Seeing From Clients Who Use ChatGPT?

Honestly, it’s a mixed bag. Some people’s ChatGPT therapy sessions help them get to a real one. Some of it worries me.

We’ve seen people clearly use AI to help write their intake forms. They’ve had a lot of back and forth with a chatbot, and it helped them put into words why they’re looking for support. That can actually be useful.

I’ve also seen AI encourage people to come into therapy. That’s a good outcome.

And for somebody with no emotional support at all, I can see the argument for having something. I just think, as a therapist, you have to be very wary of what comes down the road. How were these tools trained? How were they tested? All of this is happening so fast, and the focus is on scale, scale, scale instead of having guardrails properly in place.

Even the apps that say they’re clinically built raise questions for me. Are they biased toward a certain population or need? What are their blind spots? Because inevitably they have some.

What Does a Therapist Notice That a Chatbot Misses?

A lot. Body language, the feeling of somebody in the room, what they say, what they leave out. A chatbot only gets the words you type.

Your body pulls in data faster than your brain can. Think about walking into a college classroom where everyone is taking a test you didn’t know about. Your body feels this whoosh, that dropping of your stomach, before your brain has caught up. It just knows something’s wrong.

That same processing happens in a therapy room, for the client and the therapist. All of it is data, and it’s all important.

Here’s a general example of what that can look like. A child sits on their hands, and their body language points to something they’re trying to hold in. That could mean a lot of different things, including abuse. You don’t know until you’re in the room with them and gather enough data from what they say, what they don’t say and what their body is telling you. There’s often incongruence, and robots miss it.

Or someone comes in for anxiety, and over time their body language points to domestic violence at home. It’s a thing they don’t want to talk about. Maybe they’re part of a couple, and that’s part of the dynamic. Words alone can miss it.

Then there’s the relationship itself. A good therapist challenges your thinking. You build trust and rapport. Researchers call this the therapeutic alliance, and a 2018 meta-analysis of 295 studies with more than 30,000 clients found that a stronger alliance goes along with better outcomes in therapy (Flückiger et al., 2018).

A human therapist is also accountable under a license. We go through consultation, supervision and continuing education. We keep what you say confidential. And there’s that human relatability: seeing a real person, who may have been through something like it themselves.

Why Does ChatGPT Agree With Everything I Say?

Because chatbots tend to be sycophantic. They’re built to keep you engaged, and agreeing with you is an easy way to do it. A therapist’s job is sometimes to gently disagree.

This isn’t only my opinion. In April 2025, OpenAI rolled back a ChatGPT update after the company said it had made the model “overly supportive but disingenuous” (OpenAI, 2025). OpenAI said the update leaned too hard on short-term thumbs-up feedback.

A Stanford study published in 2025 tested popular therapy chatbots. When the researchers gave them conversations involving suicidal thoughts or delusions, the bots often went along with the user instead of pushing back, and one failed to recognize suicidal intent. The bots also showed more stigma toward conditions like schizophrenia and alcohol dependence than toward depression (Stanford HAI, 2025).

I also think about the business model. I worry that these companies are focused on getting you to use their product more, not less. Whether it’s AI or a social media algorithm, it’s built to grow your interaction with it. That’s the opposite of what therapy aims for. A good therapist is trying to work themselves out of a job.

Is What You Tell ChatGPT Private?

Mostly, no. Your chats with a general chatbot don’t have the legal protection your conversations with a therapist have.

OpenAI’s CEO, Sam Altman, said this himself in July 2025. On a podcast, he noted that people use ChatGPT like a therapist. Those conversations lack the legal privilege that covers talks with a therapist, doctor or lawyer, he said, and OpenAI could be required to produce them in a lawsuit (TechCrunch, 2025).

That’s already happened in at least one case. In the New York Times’ copyright lawsuit against OpenAI, a federal judge ordered OpenAI in May 2025 to preserve ChatGPT logs it would otherwise have deleted. The judge ended that requirement going forward in fall 2025, but logs already saved under the order stayed available (Engadget, 2025).

So my question is, how could the data you give AI be used against you? What if your conversations come up in litigation, whether that’s a divorce, an accident or something else?

Here are some other things I worry about. I’m not saying any of them is happening today. These are questions I’d want answered before I shared anything personal:

• Will you start seeing books about depression and anxiety showing up in your Amazon feed, or products marketed to you online?
• Could conversations with AI affect whether you’re approved for life insurance?
• Who knows, down the road maybe it even affects health insurance rates, depending on how health insurance continues to work. That’s an edge case, but you never know.
• What will the people marketing to you learn about you?

Some of these companies offer their tools cheap right now. I think part of the reason is that the data can be monetized later.

There’s a quieter question too. In China, if you use the chatbots there, you can quickly see they don’t speak ill of the party in power. Reporters at CBC and the Associated Press found that the Chinese chatbot DeepSeek declined to answer questions about Tiananmen Square and backed off questions about Taiwan (CBC News, 2025). Not to sound like a conspiracy theorist, but what guardrails are built into the tools we use that we don’t know about? If we treat a chatbot like an open brain, is it nudging us toward a solution somebody wants us to reach?

How Can AI Help Alongside Therapy?

Used carefully, it can be a decent assistant. Think of it as a place to practice and organize, not a place to get treated.

Some ways it can help:

• Journaling. Writing about what’s going on is helpful, with or without AI.
• Organizing your thoughts. AI can take a jumble of notes and put them in a more organized order before a session.
• Practicing a skill. You could rehearse DEAR MAN, a DBT skill for asking for what you need, before a hard conversation.
• Noticing thought patterns. If you’re learning CBT, a chatbot can help you spot the patterns in your thoughts that you’re working on with your therapist.

There’s also a separate category that people mix up with chatbots. Some digital tools have real research behind them and FDA clearance. Rejoyn, for example, is a prescription smartphone app cleared in 2024 for adults with major depressive disorder. It’s used alongside antidepressant medication, as part of care from a clinician, over a six-week program (Drug Topics, 2024). Those aren’t things like ChatGPT. As of November 2025, none of the AI-enabled medical devices FDA had authorized were for mental health uses (Psychiatric Times, 2025).

If You Use ChatGPT for Emotional Support Anyway, How Do You Protect Yourself?

Know the risks, think hard about what you share, and keep a real person somewhere in the picture.

There are going to be people who use ChatGPT, Grok or some other chatbot for this. If that’s you:

• Think very carefully about what you share. Leave out what you wouldn’t want to come back in a lawsuit, a life insurance application or a marketing profile.
• Remember what it’s designed for. It’s trying to keep you talking.
• Don’t use it in a crisis. The APA’s 2025 advisory says these tools’ ability to safely guide someone in a crisis is limited and unpredictable (APA, 2025). Call or text 988 instead.
• Look at alternatives. There are a ton of peer support groups and communities, many of them online, that can be quite good and supportive.

Two places to start: NAMI runs free, peer-led groups, including NAMI Connection for adults living with a mental health condition and a group for family members. The Depression and Bipolar Support Alliance runs free online support groups for people with depression or bipolar disorder, plus groups for family and friends.

If cost is what’s keeping you from a therapist, our guide to affordable therapy in Austin has options.

What Should Parents Know About Teens and AI Chatbots?

Pay attention to what technology your kids are using and how they’re using it. Some of the most serious harms on record involve kids and teens.

Here’s some of what’s been reported:

• Lawsuits over teen suicides. The parents of Adam Raine, a 16-year-old in California, sued OpenAI in August 2025, alleging ChatGPT contributed to their son’s death. OpenAI denies the claims, and seven more lawsuits with similar claims were filed that November (NBC, 2025). In January 2026, Google and Character.AI agreed to settle a Florida mother’s lawsuit. She alleged a Character.AI chatbot drew her 14-year-old son into an abusive relationship before his suicide. Suits from families in Colorado, New York and Texas were part of the settlement too (AP via Marietta Daily Journal, 2026).
• Sexualized chats with minors. In August 2025, Reuters reported that Meta’s internal guidelines had allowed its chatbots to engage children in “romantic or sensual” conversations. Meta said those parts had been removed (TechCrunch, 2025).
• A federal inquiry. In September 2025, the Federal Trade Commission ordered seven companies, including OpenAI, Meta, Alphabet, Snap, xAI and Character.AI, to explain how they test and monitor companion chatbots for harm to children and teens (FTC, 2025).
• Psychosis and delusions. OpenAI itself estimated in October 2025 that about 0.07% of ChatGPT users in a given week show possible signs of psychosis or mania (OpenAI, 2025). That number doesn’t show the chatbot caused anything. With hundreds of millions of users, though, it’s a lot of people in a fragile state talking to a tool that tends to agree with them.

The APA’s 2025 advisory on AI and adolescents says some young people have already formed “unhealthy and even dangerous ‘relationships’ with chatbots,” and that teens are less likely than adults to question what a bot tells them (APA, 2025).

What I’d watch for:

• Retreating. Spending more time alone with a device and less with people.
• Failing. Grades dropping, or losing interest in things they used to care about.
• Pulling away from peers. Kids need to be connecting with friends and trusted adults. A chatbot can’t stand in for that.

If you notice those signs, make sure your kid is getting appropriate support and that someone is assessing what’s going on. Ask what apps they use and what they talk to them about, with curiosity rather than a lecture. Our post on how to connect with your teen has ideas for those conversations, and our teen counseling and child counseling pages explain how we work with younger clients.

Does Just Mind Use AI?

Not in your care. To be transparent: Just Mind doesn’t use AI note-takers, AI scribes, or AI in scheduling or intake. When you reach out, a real person on our client care team reads what you share and matches you with a therapist.

Do I personally use AI? Yes, absolutely. It helps me with website maintenance and with looking at our financials by building dashboards, though we also have a fractional CFO we consult with. It’s also helped me build an app. Do I use it for my own therapy? No.

We’ve run into one AI problem directly. We saw bots trying to get around the check on our online form that confirms a real person is filling it out, and we switched verification tools because of it. That bothered me, because we require a person’s consent for certain things. If a bot fills out the form, who’s really consenting? Is that person even reading it?

Insurance is where I see AI used the most. My view is that every insurance company is using AI on claims, and increasingly we’re seeing it used to deny care. I don’t think clients would want all of their records gone through by AI, and I don’t see where they’ve transparently consented to it. I’m not alone in the worry: in a February 2025 American Medical Association survey, 61% of physicians said they’re concerned that health plans’ use of AI is increasing prior authorization denials (AMA, 2025). I wrote more about insurers and AI in an earlier post on AI and mental health.

What Laws Cover AI Therapy, and What Should Change?

A few states have acted. Texas now requires health care providers to tell patients when AI is part of their care, but no Texas law bans an AI “therapist” the way Illinois and Nevada do.

Here’s what some of the 2025 laws do:

• Illinois passed the Wellness and Oversight for Psychological Resources Act, which bars using AI to provide therapy or make therapeutic decisions. Licensed professionals can still use AI for administrative and supplementary support (Illinois IDFPR, 2025).
• Nevada (AB 406) bars companies from offering or advertising AI as professional mental health care or as a therapist, and bars providers from using AI for direct patient care. Scheduling and billing are allowed (Wilson Sonsini, 2025).
• Utah (HB 452) regulates mental health chatbots. They have to disclose that you’re talking to AI, can’t sell or share your health information without consent, and can’t use what you type to target ads (Wilson Sonsini, 2025).
• California (SB 243), in effect since January 2026, requires companion chatbots to tell users they’re talking to AI and to have a protocol that refers people expressing suicidal thoughts to crisis services. For users it knows are minors, a chatbot has to remind them to take breaks and take steps to keep sexually explicit material away from them (Jones Walker, 2025).
• Texas passed HB 149, the Texas Responsible Artificial Intelligence Governance Act, effective January 2026. It requires providers of a health care service to disclose in writing when a patient interacts with AI, and bans developing or deploying AI intended to encourage self-harm. A separate law, SB 1188, requires practitioners who use AI in diagnosis to review the AI’s output and tell patients (Texas Medical Association; Akin Gump, 2025).

What I’d like to see, in Texas and nationally:

• Deeply embedded privacy protections, and more transparency in general.
• A ban on calling a bot a therapist, or licensing AI as one.
• Age limits.
• Crisis protocols.

There’s a lot of concern in Texas about how fast this is moving and what it’s doing to our kids, partly because of what we’ve already seen social media do to kids. The challenge is timing. The Texas Legislature meets in regular session every two years, and the next one starts January 12, 2027 (Texas Legislative Council). Add in election cycles, and it’s hard to make meaningful change as quickly as the technology changes.

How Do I Move From a Chatbot to a Real Therapist?

Bring what you’ve already worked out with you. If you’ve been able to open up to a chatbot, that’s a starting point, not something to hide.

Maybe take those notes into therapy and see about making the next steps. If embarrassing things came up in your conversations with AI, you could put them in a document to share with your therapist, so you don’t have to say it again. Just strip out anything you’d rather not have stored, and keep the document somewhere you control.

You can see us in person at our North and South Austin offices or online across Texas. If you want to think through one specific thing first, a single session is a planned, one-time, 55-minute appointment. When you’re ready, get matched with a therapist.

So, Can AI Replace Your Therapist?

Not today, and I’d be careful with any company that says it can. Think about AI as a plane being built in the air. The regulation and privacy protections are coming later.

My concern is that we’ll have some kind of catastrophic problem with AI before the guardrails catch up, the way we did with nuclear power. Maybe that’s a stretch. But until the rules are in place, be very careful about what you share.

FAQ

Photo by Markus Winkler on Unsplash