For two years, Molly Quinn trusted her therapist with things she hadn’t told anyone else.

So when her therapist mentioned trying an artificial intelligence tool to take notes, Quinn didn’t immediately refuse. The 31-year-old librarian from Fayetteville, Ark., asked to research it first.

She wanted to understand where her words would go — whether they would stay local or be processed somewhere in the cloud.

Shots – Health News

With therapy hard to get, people lean on AI for mental health. What are the risks?

Replaying the session in her head

The session moved on that day, but halfway through, Quinn noticed something was different.

“She wasn’t taking notes like she usually did,” Quinn says. “The iPad was just propped up.”

That’s when Quinn realized the session was being recorded.

Quinn says she froze for a bit. But then she kept talking.

It wasn’t until she walked out of her therapist’s office that the weight of it landed.

“The more I thought about it, the more I just started getting more and more sick to my stomach,” she says. “This person who I’m supposed to be able to trust with some very private and very intense emotions had just completely disregarded something I said I was not comfortable with. I felt completely violated.”

She drove home replaying the session in her head, unsure what to do next.

New tools from new companies

Across the U.S., a growing number of therapists are experimenting with artificial intelligence tools that record sessions, generate transcripts and draft clinical notes automatically. Software companies say these tools can save hours of administrative work each week.

One company, Berries, markets its platform as a way to lighten paperwork so therapists can focus more fully on their clients and have a better work-life balance in their own personal lives.

“Berries is designed to reduce administrative burden without interfering with the therapeutic experience itself,” Tal Salman, the company’s CEO, told NPR. “It supports clinicians being more present with their clients.”

When activated, the system records the conversation, transcribes it and generates a draft clinical note for the therapist to review. Therapists can edit, revise or discard the draft. They can also save it to a medical record.

Privacy rules

Salman says that audio is automatically deleted after a session and that transcripts are stored on servers in the United States that are compliant with HIPAA, the federal law that limits release of medical information. He adds that the company does not use therapy content to train its AI models.

“Therapy session content remains private and is not repurposed for training,” Salman says. “The clinician remains fully responsible for patient care and the final documentation.”

Other companies that are offering similar services to mental health therapists are SimplePractice and Blueprint. Companies generally charge fees of $19 to $99 per month, appealing to solo practitioners who are often overwhelmed by administrative tasks.

Shots – Health News

Therapy by chatbot? The promise and challenges in using AI for mental health

Research suggests patients may be open to artificial intelligence tools in therapy, but that comfort is limited. A national survey by YouGov found that Americans remain hesitant to use AI for mental health care. Only about 11% say they would be open to using it, and just 8% say they trust it. Nearly half of respondents said they are reluctant to use the technology. Many point to the lack of human understanding as a major concern, along with fears about inaccurate advice and privacy.

Those concerns extend beyond therapy itself.

A separate survey from the health research organization KFF found that about 77% of Americans are worried about how their health information would be stored and used by AI systems.

Researchers say that the gap between curiosity and trust becomes even more pronounced in mental health care, where conversations are deeply personal. And they note that even when patients are willing to try the technology, many still prefer the reassurance of a human therapist.

Kellie Owens, an assistant professor of medical ethics at NYU Grossman School of Medicine, says privacy protections like HIPAA compliance do not eliminate the possibility of breaches.

“Regardless of what protections we have in place, that doesn’t mean data can’t be breached,” Owens says. “There are plenty of systems that are fully HIPAA compliant that still experience major data breaches.”

She says the risks are especially sensitive in therapy, where patients share deeply personal information.

Meaningful consent?

Beyond the technical risks, Owens says, written consent forms alone may not be enough.

“We have a wide body of research showing that a consent form on its own does not mean a person is making an informed choice,” she says. “People scroll through them, don’t read them or feel pressured to agree.”

Owens says recordings should always be discussed directly.

“Any time you are recording a conversation, that should require a verbal conversation that a recording is taking place,” she says.

She says trust and psychological safety are foundational to effective mental health treatment.

“If patients feel that privacy has been compromised, that can do real damage to the therapeutic relationship,” Owens says.

For Molly Quinn, those risks are not abstract.

Healthcare systems and major corporations have faced repeated data breaches in recent years, and Quinn worries that adding another layer of technology increases the chance that deeply personal conversations could someday be exposed.

“We’re going to see breaches,” she says. “Maybe not tomorrow, maybe not next week. But in a few years? I think we’re going to see them. And I don’t want my therapy session to be part of that.”

An AI listener, a different experience

For some therapists, the concern isn’t just about cybersecurity.

Marisa Cohen, a couples and sex therapist in New York, says introducing AI into therapy changes the experience, even if the technology fades into the background.

“Even the presence of AI changes the therapeutic experience,” Cohen says. “Clients know or feel like something else is listening to them. That awareness can subtly alter their disclosure.”

Therapy, she says, relies on the perception of absolute privacy.

“When you introduce something that’s being stored electronically, it raises additional questions about trust and safety,” Cohen says. “It’s essentially a third party.”

Cohen also worries about accuracy.

“If errors are introduced and a clinician isn’t meticulously checking the notes, that error is now part of the record,” she says. “If those notes are ever subpoenaed, that becomes part of someone’s history.”

Treatments

The (artificial intelligence) therapist can see you now

A boon for some therapists

Some clinicians are welcoming the technology.

Kym Tolson, who brands herself online as “The Traveling Therapist,” has been practicing for 30 years and now runs a fully remote business, seeing clients in Virginia while traveling full time. She uses an AI recording and documentation system to manage what she describes as a heavy administrative load.

“Most clinicians spend about 10 hours a week on administrative tasks, and five to seven of that is documentation,” Tolson says. “With the AI system, I spend about two minutes per client where it used to take me 15 to 20.”

For therapists carrying 20 to 30 clients, she says, that difference is significant.

“It’s given me my life back,” Tolson says. “I don’t have notes following me around, haunting me. After I see my client, I review the note, sign it and I’m done.”

Tolson says the flexibility has allowed her to maintain a healthier work-life balance and spend more time focused on clients rather than paperwork. She says many clients respond positively to her mobile lifestyle, often asking where she’s working from during sessions.

Research in other parts of healthcare suggests the time savings can be substantial. In one large analysis of AI scribes at The Permanente Medical Group, clinicians saved the equivalent of nearly 16,000 hours of documentation time in one year — roughly 1,800 workdays — and reported improved satisfaction and patient interactions. The study focused on medical clinicians rather than therapists, but it highlights the administrative burden these tools aim to address.

Shots – Health News

Their teenage sons died by suicide. Now, they are sounding an alarm about AI chatbots

Living Better

Teens are having disturbing interactions with chatbots. Here’s how to lower the risks

Tolson says she discusses the recording process in detail with clients beforehand and makes clear that participation is optional. She also reviews each AI-generated note before it becomes part of the medical record.

“There are times it will hallucinate,” she says. “The clinician has to be very careful. You have to double- and triple-check.”

While she acknowledges that no digital system is entirely risk-free, Tolson says she feels as comfortable using the tool as she does using electronic health records.

Starting over with a different clinician

For Molly Quinn, the conversation about AI note-taking will always trace back to that November session, the one where she realized she was being recorded without her consent.

Afterward, she canceled her next appointment.

Her therapist said they could stop using the tool. But by then, Quinn says, something felt broken in the relationship.

“The trust was gone,” she says.

In the weeks that followed, she began searching for someone new. By January, she had found another therapist, and from the start, she made it clear she did not want AI used in her sessions.

The experience didn’t turn her away from therapy itself, but she remains firm about her boundaries.

“This is something that is a private conversation,” Quinn says. “Tech companies shouldn’t be in the room.”