Mohit hosts the Medical Centric Podcast, where most guests are doctors talking about prevention: eat well, move, manage stress. Then I showed up to talk about schizophrenia, phones and AI. He asked the questions families ask in their first year, and a few I’m still wrestling with after six.
Listening back, I noticed one thread running through all 42 minutes. In serious mental illness, the people who see the most, the families, usually have the least say in what happens next. Technology could change that. It also raises questions none of us have answered yet.
A 15-minute snapshot of a 30-day story
There is no lab test for schizophrenia. No blood draw, no scan. A diagnosis rests on a short appointment and whatever the family manages to say before time runs out. So people bounce between labels: bipolar disorder with psychosis one year, schizophrenia two years later, each label bringing a new medication and more months lost.
Mohit offered his own example. For six months his watch has flagged a heart rate below 40, about ten times a month, always in the morning. Any single clinic visit would find nothing. Six months of data tells a different story.
That’s the whole idea. Sleep, mood and symptoms tracked day over day, summarized before the visit, turn a family’s “I think something’s off” into a pattern a psychiatrist can act on. The clinician still decides. They just stop deciding from a snapshot.
Something to think about: if your observations arrived at the appointment as data instead of a story told in the hallway, would anyone listen differently?
A phone that notices
Relapse rarely comes out of nowhere. In one pilot study, a smartphone picked up behavioral changes 71% more often in the two weeks before a schizophrenia relapse than at other times. The example I gave: someone who normally sets his phone down by 9 p.m. is up all night, three nights running. With his consent, an app texts his care team: he hasn’t slept in three nights, someone needs to check in.
The technology to send that text exists today. What most care teams don’t have is a person assigned to read it. Until they do, families remain the alert system.
The flood: agreeable chatbots and untested apps
Mohit asked whether AI was my main focus. It’s one part. AI that sorts through data to help a clinician is one thing. A chatbot talking to someone in psychosis is another. These tools are built to be agreeable. OpenAI itself rolled back an update in 2025 for being too sycophantic, and has since described new safeguards for people in crisis. Safeguards help. But a tool designed to agree is a risky companion for a mind that needs its beliefs gently tested, not confirmed.
Apps raise a quieter problem. Half of mental health apps vanish within three years, and only a small share of the rest have been tested in any formal way. An untested mood tracker or brain game may not hurt anyone. But “probably harmless” isn’t the same as “will help,” and an app that asks for location and sleep data deserves a hard look at who’s collecting it.
Not just crisis: rehearsal
The part of the conversation that gave me the most hope was about moving forward. In a small trial, people with schizophrenia who practiced job interviews in virtual reality were more likely to have a job offer six months later. Picture someone stable after ten years of illness who wants to date again and has no idea how to explain the last decade. VR lets him rehearse that conversation until the words come out right.
I think about technology in three stages: stopping a crisis, managing the illness day to day, and building a life again. There are tools for every one of them.
The question I couldn’t answer
Many people with schizophrenia or bipolar disorder don’t know they’re sick. It’s called anosognosia, and it’s a symptom of the illness, not stubbornness. Now imagine your adult son, just home from the hospital, insisting he’s fine. You want an app that would show his sleep falling apart before the next crisis. He refuses. If psychosis tells him he’s being watched, the app proves it.
Advocates for civil rights are right that he has the right to say no. Families are right that “no” can end in hospitalization, homelessness or worse. So are you saving his life, or invading his privacy? I told Mohit I don’t have a clean answer. I still don’t.
If your family has faced this choice, I’d like to hear how you handled it. Just hit reply.
When the doctor’s office is too much
For two to three years, my daughter could not sit across from a doctor. If she looked at someone’s face, her voices turned it into screaming. Video visits were the only way she could be treated. She kept her eyes down the whole time, but she was there. For families in rural areas, or anyone searching for a clinician who shares their language or culture, telepsychiatry can be the only door that opens, though the rules across state lines vary.
Where we ended
Mohit’s last question was about prevention. Could better diet, exercise and less stress prevent these illnesses? They help, and researchers are even studying whether a ketogenic diet can help with psychosis, though that work is early. But serious mental illnesses are brain illnesses with strong genetic roots. Nobody gets schizophrenia because they ate the wrong thing or didn’t exercise enough.
Which is how we closed: people who develop these illnesses didn’t do anything wrong. Neither did you.
Watch the full conversation here, or jump to any chapter linked above.
Everything we discussed is covered in depth in my book, Connected Care: A Practical Guide to Technology for Serious Mental Illness. Get it on Amazon.
