AI Is Already in the Therapy Room. The Question Is Who's Holding the Reins.
Patients brought AI to therapy before anyone asked them to. The states are legislating and the clinicians are alarmed. The real opportunity is the one few people are arguing about.
The American Psychological Association surveyed more than 1,200 licensed psychologists this spring. The profession it found was not debating whether AI would enter practice. AI already had. Seventy-seven percent said their patients had used AI for support. More than a third said patients treated a chatbot as an additional mental health professional. Thirteen percent said a patient had formed an "intimate or relationship-like" connection with one.
The debate about AI in mental health is a status report, not a forecast. Patients did not wait for a pilot or a clinician's blessing. They opened an app between one Tuesday session and the next, and they started talking.
That gap, the six days and twenty-three hours when no clinician is in the room, is the whole story this week. Every thread I have been following traces back to it.
Start with the alarm. In the same survey, 93% of psychologists had concerns about certain patients using chatbots, 97% worried the tools could reinforce dysfunctional beliefs, and 94% doubted today's chatbots can treat conditions with appropriate nuance. Researchers have a name for the failure mode: the "sycophancy trap," the tendency of an engagement-optimized model to agree and validate rather than challenge. Part of a therapist's job is to disagree with you, gently. A product built to maximize your next message rarely will. Only 24% of psychologists think patients will one day prefer chatbots to humans. This is not a profession in retreat. It is a profession watching an unsupervised tool do supervised work.
Then there is the law, and it is arriving fast. Illinois' Wellness and Oversight for Psychological Resources Act forbids AI from delivering therapy on its own, or being marketed as a therapist, unless a licensed professional stays in charge, with fines up to $10,000. Nevada bars AI from standing in for a counselor, including in schools. As of July 1, Tennessee prohibits AI systems from representing themselves as licensed mental health professionals and treats violations as deceptive practices. Utah requires mental health chatbots to disclose that they are software and limits how they use patient data. The mechanisms differ. The principle does not: a licensed human must remain accountable for the care.
The thread most people miss is the one without a scandal in it. While patient-facing chatbots absorb the headlines, AI has been entering therapy through a second door, on the clinician's side. A category of documentation tools now drafts session notes, carries treatment goals forward, and links one session to the next so that progress stays visible across weeks of care. The framing in that corner of the industry is more sober. As one vendor's 2026 review put it, the value of documentation "lies in maintaining continuity, supporting reflection, and making progress visible" over time. That is AI handing the therapist a better memory.
Even here, caution is earned. A 2026 review of AI note-taking tools found meaningful quality gaps between machine-generated documentation and notes written by clinicians. The lesson is not to avoid AI. The lesson is not to use it unsupervised, which is what the legislatures said too.
Put the three threads side by side. Patients are pouring into the between-session gap because it is lonely and a chatbot is awake at 2 a.m. Clinicians are alarmed because the tool filling that gap is optimized for agreement. Regulators keep drawing one line: a licensed human stays responsible. What is being sorted out is who AI should be pointed at. The patient, alone, or the clinician, better informed.
I think the answer is clear, and I am not a neutral observer. At Lisner we build for exactly that gap. The clinician sees what a patient chooses to share from the week, so a session can start with context instead of reconstruction. Therapy depends on the alliance, the trust and challenge between two people, which is also why the sycophancy trap is dangerous. AI can strengthen that relationship or substitute a frictionless imitation for it. Those are very different products. The survey data, the statutes, and the documentation market are each, in their own language, voting for the first one.
Patients have told us what they need: someone paying attention between sessions. A chatbot with no license and little reason to say something you do not want to hear is the wrong someone. The clinician is the right one. The job is to make sure the clinician finally can.
