Therapists Are Asking Patients About Their AI Chatbot Conversations

Suzi Sanford types into her computer from a cozy nook in her bedroom. She's not texting a friend. She's talking to Claude, the generative artificial intelligence chatbot owned by Anthropic. The 32-year-old Fitchburg mom uses the platform to process emotions between therapy sessions, asking the bot to summarize her recent job and relationship stresses.

The bot instantly spit out a summary of themes: feeling "unseen" and "unheard" in various life situations. Sanford says AI helps her piece together "fractured" thoughts when emotional upheaval strikes. It's a growing pattern across the country.

According to a recent poll from KFF, 16% of adults turned to AI tools and chatbots in the past year for their mental health. The percentages skew higher for younger adults. A Pew Research Center survey found two-thirds of teenagers interact with chatbots, with 28% firing off messages every day. Among people with mental health conditions who've tapped into AI tools, nearly half do so for psychological help.

As more patients bring these conversations into the therapy room, clinicians are adapting. WBUR's reporting documents how mental health providers are now routinely asking clients about their AI use.

Dr. Christine Crawford, a Boston psychiatrist, consults AI to process emotions that come up in her own work. In her Back Bay office, she opened the ChatGPT app on her phone and demonstrated how she's turned to it to work through difficult patient sessions. She described a recent case involving a patient who experienced horrific childhood trauma.

A soothing male voice responded with validation and perspective. But Crawford said conversing with the chatbot "freaks" her out — the experience "feels too real." So she usually types. She likened this use of AI to the way she'd turn to a supervisor during medical training or a peer psychiatrist in hospital work. Now in private practice, it's easier to reach for her phone.

She stressed that she doesn't share any patient information with chatbots. Like many clinicians, she's concerned about data privacy on AI platforms (a problem that has plagued users for years, frankly). She asks every patient about their AI use and how it makes them feel.

"All of that allows me to get a better sense as to what's inside of my patient's brain," Crawford said. "To use their conversations off of chat as a jumping off point for better connecting and better engaging with the patient in person."

Some researchers argue it's "essential" for clinicians to ask clients about their AI use. A recent paper in the journal JAMA Psychiatry called for providers to "approach these conversations with curiosity rather than judgment."

Clinical social worker Dan Sutelman, of Newton, said he talks about AI and its impacts when his clients bring it up; he estimated about 20% do. He even asks some kids who use the tech to take out their phones and communicate with ChatGPT during sessions if they're having trouble talking to him about something.

"I really would be listening to the child to see what is it that they're trying to get from it, whether it's information or an emotional connection, guidance or validation ... or just the sense of being accompanied," Sutelman said. "But I also would be listening for how are they relating to the chatbot relative to how they're relating to me."

Some adult clients have told Sutelman they asked a chatbot about observations he's made during therapy. He said that can undermine the treatment, due to the pattern of how AI responds. "It ends up sort of invalidating the work that I'm doing with the person and validating something about the person," he said.

Generative AI chatbots are trained to use supportive, empathetic language. They're also designed to keep a person engaged, often with lots of questions. Validation can be good, some mental health clinicians said, but they also warned it can be unhealthy or risky when it replaces human accountability.

The physical reality of this interaction matters. Sanford types on her keyboard in her bedroom. Crawford taps her phone screen in her office. Sutelman watches a teenager's thumbs move across a glass display. These aren't abstract conversations. They're happening in real spaces, on real devices, with real data flowing somewhere.

Whether patients actually benefit long-term remains the real question.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn