Technology news for families navigating serious mental illness — translated so you don’t have to.
This week, the rest of the world caught up to a chapter I’ve been wanting to write for a while.
A new survey found that more than three out of four psychologists now say their patients are turning to AI chatbots for mental health support. The researchers gave it a name: the sycophancy trap. AI chatbots are built to be agreeable. They mirror whatever you tell them. For most people, most of the time, that feels comforting. But if your loved one lives with psychosis, paranoia, or delusional thinking, an AI that’s built to agree with whatever it’s told isn’t comforting — it’s dangerous. It doesn’t say “that doesn’t sound right.” It doesn’t know the difference between supporting someone and reinforcing a belief that isn’t true.
This is exactly what I flagged in a recent post, before this data existed. It’s good to see the clinical world catching up with words and numbers.
This week’s top story
Psychologists warn of a sycophancy trap as patients increasingly turn to AI chatbots for therapy
PsyPost — June 18, 2026
It’s not just clinicians raising the alarm. Vermont just became the first state to make it illegal for an AI system to act as an independent mental health provider — no diagnoses, no treatment plans, no therapeutic guidance, unless a licensed professional is reviewing it. The bill’s sponsor put it simply: “These tools are not healthcare.” Vermont isn’t saying families can’t use AI tools at all — administrative support, symptom tracking, and clinician-supervised tools are still fine. It’s drawing a line around AI acting alone, in place of a person who’s trained for this.
A couple of smaller stories worth knowing about, if you want to go deeper:
- TechTarget reported that patients are finding mental health chatbots judgmental in some cases — the opposite failure mode, and a reminder that these tools are still inconsistent, not just oversimplified.
- The American Psychological Association published new guidance for clinicians on how to actually talk to patients about their AI use — worth a look if you want language for a similar conversation with your own family.
None of this changes what’s in Connected Care. It confirms it. The tools that matter are the ones built and studied specifically for this community, with real oversight behind them — not an open-ended chatbot standing in for a person who’s trained for this.
Until next week,
Nicole
