New survey: 85% of Americans have hesitations about AI mental health advice

There’s a version of the future where you never wait for a therapist again. Feeling low at midnight? A chatbot is standing by—endlessly patient, free, and fluent in every coping skill ever published. It sounds like an appealing pitch, but most Americans aren’t so sure they want AI in their mental healthcare.

In Thriveworks’ 2026 Pulse on Mental Health Report—our third annual nationally representative survey of 2,000 U.S. adults, conducted by Wakefield Research in March 2026—85% of people said they have at least one hesitation about AI mental health advice. Some even say the increasing use of AI has created a new source of anxiety rather than being a tool to ease it. But it’s not a simple “no thanks” across the board. Opinions on AI range widely across generations and genders, depending on the situation.

We asked three clinical experts to help us understand what the data reveals about what people really want from therapy, what AI can and can’t replicate, and what responsible use of AI to fill gaps in mental healthcare actually looks like.

Most Americans are skeptical about AI in therapy. For some, it’s become an active stressor

The vast majority (85%) of Americans have at least one hesitation about using AI for mental health advice, and 64% have less confidence in AI than they do in human therapists.

The most common concern, cited by 45% of Americans, is that AI can’t understand their unique situation. More than one in three people (37%) said they want the experience of talking things through, rather than just being handed a recommendation from an AI bot. And 25% worry that AI can’t handle a crisis or suicidal thoughts safely.

Beyond just being a concern, the increasing use of AI in and out of the therapy room has become a new source of anxiety for roughly 1 in 4 Americans (21%).

There’s a generational and gender divide about if and how AI should be used in therapy

Skepticism toward AI in mental health care drops steadily with each younger generation. Among Boomers, 69% say AI should have no role in therapy, while 45% don’t even want it handling administrative tasks for their providers. Gen Z is the only generation where openness exceeds skepticism: only 34% want AI out of therapy entirely, and 41% are open to it as an alternative when therapy isn’t available.

Millennials are pretty conflicted about this. They’re the most likely of any generation to say AI access has made them more anxious (23% of Millennials, versus 16% of Gen Z, 11% of Gen X, and 7% of Boomers), and they’re the most likely to say AI can be part of the process as long as it doesn’t replace the human (34% of Millennials, versus 25% of Gen Z, 26% of Gen X, and 23% of Boomers).

There’s a gender divide at play as well. Women are more resistant to AI in mental healthcare, with 35% saying it should play no part, compared to 24% of men. Men see more potential in AI for mental health: They’re more likely than women to see AI as being just as effective as a human therapist (13% versus 8%), and 40% are as confident or more confident in AI than in a therapist, compared to 32% of women.

Generally, Americans see AI as a tool, not a therapist

For all the skepticism, outright rejection is a minority position. Only 30% of Americans want AI to have no part in mental health care at all. The rest are open to it playing some role: 21% would only want it used as an administrative tool, while 27% say AI can be a part of the therapeutic process as long as it’s not replacing the human element.

This tracks with how survey respondents are already integrating AI into their mental health toolkit. Of those who have used AI for mental health support, three out of four (75%) used it as a supplement to therapy rather than a replacement. And 35% of them found it to be as effective as human care—an interesting number to keep an eye on, and one we asked our clinicians to dig into below.

What this tells us about the role of therapy in our lives—now and into the future

Looking at the top hesitations around AI mental health advice, it’s clear that most Americans value the care and nuance that comes with traditional therapy. Patients want to be understood—not just in the sycophantic way that AI bots tend towards, but in a deeply human one. They want the experience of talking things through, rather than just being handed a recommendation.

“It’s super easy to type into a chatbot and to get an answer back, but there’s not that rapport,” says Hallie Kritsas, licensed mental health counselor at Thriveworks.

Providers agree that so much of the progress that happens in therapy relies on that human-to-human dynamic. “A huge part of our work is picking up nonverbal cues, which a chatbot doesn’t see,” says Jami Dumler, licensed clinical social worker and clinical programs director at Thriveworks. “Personally, I’ve seen so many breakthroughs happen in therapy because of a nonverbal cue or a silence or a small bit of emotion that I can pick up on with the client in the room.”

Patients and providers alike also worry that AI can’t handle a crisis or suicidal thoughts. “My job is to keep you safe,” explains Kate Hanselman, psychiatric mental health nurse practitioner and VP of clinical psychiatry at Thriveworks. “AI doesn’t have an ethical, moral, legal, or financial imperative to catch something like that and help. AI doesn’t have consequences. Which is really, from a clinical perspective, very scary.”

How AI might help bridge the gap to care

For all the variation in how Americans feel about AI, the clinicians who actually sit with patients don’t seem to be concerned about AI deterring people from therapy. If anything, it may help some of them get there.

The gender gap may be the most obvious example of AI opening a door instead of closing one. Men are more open to AI across the board, which Dumler sees as a low-stigma way in. “If someone can access therapeutic tools in any way, I think that’s a start and a win,” she says. “We’ve also seen more men coming to therapy than we ever have. So I think we’re just seeing an overall positive trend in men being open to mental health support, regardless of what type of tool it is.”

Gen Z may be the most AI-open generation, but Dumler doesn’t read that as anyone swapping the therapist for the chatbot. Plenty of them still seek traditional therapy, she says—they’re just “more open to AI tools between sessions or using AI to source extra information outside of therapy.”

And, for some people just looking for basic coping skills, AI can provide that, particularly if you’re not dealing with any other complex mental health issues. “If someone generally has had a good upbringing, limited trauma, overall a stable ability to cope, they might just need an extra little buffer,” Dumler says.

“But if you’re someone that has a lot more challenges on your plate, you probably need more complex support like a live therapist,” she adds. “Particularly if we’re talking about severe mental health or really established mood disorders like bipolar, major depression, generalized anxiety, AI is not going to be able to meet that gap.”

What AI as a supplement to therapy actually looks like

Reducing administrative burden so clinicians can focus on their patients

When it comes to the ways clinicians are actually putting AI to work, the AI scribe is a clear favorite. These tools sit in on a session and record, transcribe, and draft the clinical note—always with a patient’s informed consent.

Even though a therapist still reviews these notes for accuracy, an AI scribe can take this from a five- to 20-minute task down to a two- to three-minute one. That reclaimed time lets clinicians be more present in sessions, take a break between sessions so they can show up fresh for the next patient, and fit more patient-facing work into their schedule. “It really opens the door for more care coordination, therapist self-care, or even an extra session or two,” Dumler says.

Creating personalized and interactive practice tools

Thriveworks clinicians also shared that AI can help create expert-informed tools to help their patients practice skills and get more out of therapy.

“Back in the day, I would give you a worksheet. Now you can practice with a chatbot,” says Hanselman, who has seen tools like internal family systems or CBT chatbots that may help skills stick better than a static worksheet. Dumler has also used AI tools to create custom meditations based on patient goals.

Even something as simple as a follow-up recap of a session generated by an AI scribe can be valuable. “That bridge between sessions can really help patients remember that hour and keep their goals in the forefront,” Dumler says.

Being available when a therapist can’t (as long as you aren’t in crisis)

One of the major benefits of AI is that it can be used any time, anywhere. “I can’t be available to you 24/7, and when things feel really hard in the middle of the night, if you live alone and don’t have a partner or a friend you can call, AI can help people feel less alone,” Hanselman says.

All three clinicians draw a hard line when it comes to crisis management though, and make sure to emphasize this with their patients. “The conversation I have with everybody is: If you are in danger, if not being here starts sounding good, if there’s any kind of risk present, I want you to talk to a human being,” Hanselman says. Dumler agrees that she works with all patients to understand the signs of escalation, noting that 988 (the suicide prevention line), 911, or their local emergency room should be the first line of defense.

Kritsas asks her patients who use AI between sessions to bring it back to the therapy room so they can examine it together. “I would never tell somebody that they shouldn’t use some sort of resource in between counseling sessions, but we do talk about why: What was going on? What did you ask them? What did they say? Is that something that we can actually implement in your life or can we talk about other things as somebody who intrinsically knows you as a person?”

Informed consent and HIPAA-compliant tools

Underneath every responsible use case for AI in mental healthcare is the same non-negotiable: informed consent by the patient.

“It’s our duty as therapists and agencies in mental health to make sure a patient knows exactly what AI we’re using, what it’s used for, their security and privacy measures—and then consenting fully to it,” Dumler says.

She adds that responsible providers should always be using vetted, safe, HIPAA-compliant vendors for AI use, with a Business Associate Agreement (BAA) in place to protect private patient health information. “There are probably therapists that just grab the latest AI tool online and haven’t had that BAA agreement signed, don’t know exactly how the data is being stored or utilized, and are not having conversations with their patients about consent. That would be a big red flag to me.”

If you’re a patient who’s unsure about how your provider may be using AI, Dumler encourages you to ask. And if you ever feel uncomfortable, remember that consent can be revoked at any time. “AI does not have to be used in the therapy room.