The Chatbot-Therapist Era Is Ending. What Replaces It Matters More.

Last November, a Harvard psychiatrist told Congress that OpenAI had itself reported over one million users a week having ChatGPT conversations with explicit indicators of potential suicide planning. One million a week, on one platform.

I keep returning to that number. This week made clear that three groups who rarely agree, psychiatrists, state legislators, and venture capitalists, have arrived at the same conclusion about what to do about it. The chatbot-as-therapist experiment is being dismantled in real time. The question now is which role AI should have in mental health care.

Start with the clinicians. A Medical Daily piece this week pulled together the research record from the first half of 2026, and it reads like a verdict. A Brown University study from March identified fifteen distinct ethical risks in AI-generated mental health responses, including something the researchers called "deceptive empathy," where a chatbot mimics the language of care without the clinical understanding to apply it. Stanford HAI research found chatbots failed to respond safely in crisis scenarios roughly 20 percent of the time, against 7 percent for human therapists. In a crisis, that gap is the whole ballgame. The Psychiatric Times review cited in the same piece concluded chatbots should be considered contraindicated, a word clinicians do not use lightly, for users who are actively suicidal, psychotic, or in acute crisis.

The statehouses heard the same signal. As of June, eleven states have passed chatbot laws, with a twelfth awaiting a governor's signature, according to an IAPP analysis by ZwillGen attorneys. The specificity strikes me more than the volume. Six states now prohibit operators from letting a chatbot represent that it provides professional mental or behavioral healthcare. Several require evidence-based methods for detecting suicidal ideation, which rules out keyword-filter theater. Illinois had already banned AI therapy outright in 2025. The wave has been unusually bipartisan, moving through Republican- and Democratic-led legislatures alike. When Springfield and Sacramento agree, something real is happening.

Then there is the money, often the most honest signal. A research2guidance analysis of 2026 funding found capital concentrating hard. Five rounds above $50 million absorbed roughly 74 percent of all mental health AI capital in 2025 and 2026, and nearly all of it went to one shape of company: psychiatry copilots, clinical AI assistants, and hybrid clinician-AI platforms. Their line on the losing category was blunt: "wrappers around public LLMs no longer survive diligence." Their line on the winning one belongs on a pitch deck: "human in the loop is now a precondition, not a feature."

Even the good news for AI this week came with the same asterisk. A multi-institutional randomized controlled trial published in NEJM AI on July 22 tested Flourish, a generative-AI well-being app, with 486 university students. The trial found greater positive affect, resilience, and social well-being. It found no significant effects on depression, anxiety, or stress. Even in a careful, preregistered trial, the app did not move the clinical measures. The ceiling is where clinicians said it would be.

That leaves the question of where AI should sit, if not in the therapist's chair.

The most compelling answer I read this week came from Harvard Medicine Magazine's feature on John Torous, the psychiatrist who gave that congressional testimony. His argument is that the future of AI in this field is making sense of the context around a patient: sleep, movement, and the texture of daily life between appointments. He offers a thought experiment I cannot improve on. Imagine someone writing a goodbye letter to colleagues. Is it a planned retirement or a suicide note? A chatbot alone does not know. A clinician with real-time context might. He reports that in his lab's work, feeding this kind of "digital phenotyping" data into a frontier model let it flag worsening depression in every one of the mock cases.

That reframing matters. The failure mode of the last three years was treating the 50 minutes a patient spends in session as the whole of care, and outsourcing the other 10,030 minutes of the week to an engagement-optimized chatbot. The opportunity is the opposite. Keep the qualified human at the center of care, and use AI to close the visibility gap, so the therapist knows what the week looked like before the session starts.

I run a company built on that premise, so discount my conviction accordingly. This week I did not need to make the argument myself. The psychiatrists made it in the literature, the legislators in statute, the investors with term sheets. The era of asking whether AI can be a therapist is closing. The harder question, how AI makes a therapist better informed between the hours they are in the room, is the one worth working on.