AI Chatbots Fill Therapist Shortage Gap, Raising Legal Questions

A Reddit user recently shared a screenshot showing a ChatGPT conversation that lasted nearly 24 hours. The person was going through difficult times, and the responses from other users revealed a pattern: many described their own extended sessions with AI chatbots following breakups, job losses, and sleepless nights.

This trend reflects a growing reality in mental health care. With roughly half of psychologists unable to accept new patients and more than a third rejecting insurance entirely, according to American Psychological Association data, people are increasingly turning to AI for support that human providers cannot offer.

Why it matters

The shift to AI mental health support isn't driven by preference—it's driven by access. Uninsured adults are more than twice as likely as insured adults to use AI for mental health advice (30% versus 14%), according to March polling from KFF. This creates a two-tier system where those who can afford traditional therapy continue seeing human clinicians, while others rely on chatbots that operate in a legal gray zone with no clear accountability when things go wrong.

Legal liability remains unsettled

The question of who bears responsibility when AI chatbots provide harmful mental health advice is now playing out in courtrooms. Families of teenagers who died by suicide have filed lawsuits against Character.AI and OpenAI, framing their claims as product liability cases rather than traditional media lawsuits.

When Character.AI and Google attempted to dismiss one case, a judge rejected their motion. Both companies have since agreed to mediation with families in Florida and Colorado. A separate case against OpenAI continues through the legal system.

Section 230, the law that shields tech platforms from liability for user-generated content, may not apply to chatbots because they generate responses rather than simply hosting content from others. This distinction could fundamentally change how courts treat AI companies.

States move to regulate

State governments are implementing restrictions without waiting for federal action. Illinois, Nevada, and Utah passed laws this year prohibiting AI from making therapeutic decisions. Texas opened an investigation into chatbot marketing practices, and New York now requires bots to recognize self-harm indicators and direct users to professional resources.

Pennsylvania took a different approach, suing Character.AI in May for allegedly operating a chatbot that posed as a licensed psychiatrist, complete with a fabricated medical license number. The state argues this constitutes practicing medicine without a license.

Where chatbots fail

Research shows chatbots can replicate basic therapeutic language patterns effectively. In some blind tests, they've been rated as more careful than human clinicians in their responses. However, Stanford researchers identified a critical flaw: when presented with subtle suicidal ideation—such as someone mentioning a lost job and then asking about tall bridge heights—chatbots provided the requested information rather than recognizing the warning signs.

Psychiatric researchers describe an "amplification spiral" where chatbots' agreeableness, mirroring behavior, and personalized responses create an illusion of understanding. This design makes the technology particularly risky for people in crisis who lack access to human alternatives.

These details were first reported by Quartz.