Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Millions of Americans make a choice between AI or nothing

Access to mental healthcare needs attention.

[Photo: Getty Images]

Profile photo of Sarah Baldry

Chief Marketing Officer · Wysa

The American Psychological Association (APA) has issued a new advisory urging the public to be cautious about using AI chatbots for mental health support. The warning is necessary, as generative AI should not be treated as psychotherapy, psychological treatment, diagnosis, or crisis care. The trouble is that for millions of people, the alternative is no support at all.

In a new survey, the APA found that AI use has rapidly become part of the therapy process. Of the more than 1,200 psychologists surveyed, 77% had spoken with patients who used AI for engagement, support, or other reasons.

Nearly two in five said patients had used AI to self-diagnose a mental health condition. More than a third of psychologists (36%) said they had noticed a patient developing some dependency on a chatbot, and 15% had seen patients develop distorted thinking or delusions tied to one. These figures are worrying.

A chatbot is not a therapist, and no AI tool should be allowed to muddy that distinction by being designed to make clinical judgments, diagnose conditions, or carry responsibility for a person’s care. I work at Wysa, a company that builds AI mental health support tools, so I have an interest in this debate. I also have a clear view of where the boundary belongs.

While the public needs a clear warning not to use generic AI as a therapist, many people use AI for support because access to mental health care is painfully uneven in the U.S. Even if every licensed therapist in America doubled their case numbers tomorrow, millions would still face barriers to care.

When the alternative isn’t therapy

The intended journey sounds simple: Notice you are struggling, find a qualified professional, make an appointment, and get help. In practice, it’s challenging to do so. Mental Health America has reported that one in four adults with any mental illness had an unmet need for treatment, and more than five million adults with mental illness were uninsured.

You may not even be able to get an appointment even if you try. Around 40% (137 million) of the U.S. population lives somewhere without enough mental health professionals, according to official government research. The problem is particularly acute in rural areas.

If responsible mental health organizations are forced to leave current technology spaces entirely, generic chatbots, search engines, social media feeds, and online forums could take their place. We should not be surprised when people use whatever answers first. The practical question is how technology can responsibly expand support without attempting to replace human therapy.

America needs an option between “see a therapist” and “you’re on your own.” Some people need clinical care. Others need structured and engaging self-help support before their distress escalates. Many need help understanding what they are feeling, what to try next, and when self-help is no longer enough.

The option in between

AI’s legitimate role in mental health offers transparent, evidence-based support with safety guardrails, designed and tested by clinicians, helping people take the next constructive step.

A self-help tool can be designed to support moments such as panic, poor sleep, or worry by offering structured exercises and prompts, pointing people toward the right resources, and making clear when human support is needed.

None of that requires AI to pretend to be a therapist. In fact, it requires the opposite, including clinical input, tested frameworks, transparent privacy, and the ability to escalate when risk appears.

This is where the debate deserves more precision. The APA itself notes that it is unclear whether patients are mainly using general-purpose chatbots or tools developed with clinical input and tested against established therapeutic frameworks.

A meaningful difference exists between a generic AI chatbot drifting into mental health conversation and a clinically governed tool designed for bounded self-help, with transparent limitations, risk detection, and escalation. The first should concern us deeply. The second deserves scrutiny and regulation rather than being dissolved into the same category.

What good regulation looks like

The next wave of innovating mental health technology companies will be assessed on whether they can meet the governance, evidence, and accountability standards of clinical infrastructure, regardless of how the product is positioned.

Regulatory standards are evolving too, which helps bring clarity. In the U.K., the government regulator MHRA is developing regulations around healthcare AI to set the rules for safety, classification, and monitoring. Meanwhile, the National Institute of Health and Care Excellence (NICE) evaluates clinical evidence and cost-effectiveness.

In the U.S., the FDA can recognize digital mental health tools with the Breakthrough Device program, a voluntary program for certain medical devices and products offering effective diagnosis or treatment for life-threatening or irreversibly debilitating diseases or conditions. The difference between a governed tool and a generic chatbot is relevant evidence, accountability, and design.

The APA advisory is a necessary intervention. It draws attention to real risks at a moment when AI tools are being adopted faster than society can evaluate them. The better answer is regulation.

Sarah Baldry is the chief marketing officer at Wysa.