We could all use a little help these days. Americans are lonelier than ever, and more than 1 in 5 now suffers from a mental illness. Demand for support is so great that access to quality care remains a persistent challenge, and burnout among mental health workers is at an all-time high.
Given this reality, a growing number of patients are being drawn to non-human therapists powered by AI, like ChatGPT. And while it’s encouraging to see more people recognize the importance of mental health, it’s critical we remind them that AI should serve as a supplement to human care, not a substitute.
Confronting the reality of the situation
Across age groups, but especially young people, AI tools are becoming a primary source of emotional support. They are available at any hour, respond instantly, and create a sense of privacy that traditional clinical settings have struggled to match. For someone that’s nervous or hesitant to speak to someone, that can sound pretty appealing.
A recent study found that 13% of adolescents and teenagers in the US (approximately 5.4 million individuals) have already used AI for mental health advice. Of those 5.4 million users, more than 65% engage at least monthly. What starts as curiosity becomes routine. People return to these tools, share personal details, and begin to rely on them during difficult moments. In many cases, those interactions take on the shape of a relationship.
But these tools are operating outside clinical oversight – meaning, there are no shared standards for how they respond, no consistent safeguards, and no clear accountability. Groups like the American Psychological Association and the World Health Organization (WHO) have begun to issue guidelines on AI-assisted mental health tools, but things are moving fast, and for individuals navigating anxiety, depression, or more complex conditions, a lack of professional structure introduces real risk.
Why fully automated mental health care falls short
AI is already deeply integrated into the broader healthcare industry, and the technology has helped improve workflows, offload administrative tasks, and aid in clinical decision-making. Those benefits also extend to mental health care, but there are nuances to treating the mind that cannot be optimized by speed or consistency alone.
A patient rarely presents with a simple, isolated issue. What appears as stress may be rooted in trauma. What sounds like fatigue may be linked to depression or physical health. A clinician listens for patterns, context, and meaning – sometimes gleaning more insight from the unspoken clues. And while AI can process language and support structured interactions, it’s unable to fully replicate the judgment that comes from training and experience, or the connection that develops between two people over time.
That human connection is usually where progress begins. Feeling understood, challenged, and supported by another person can change how someone engages in their own care. Fully automated systems can miss that depth, often oversimplifying complex conditions, overlooking underlying issues, and offering responses that feel helpful in the moment, but lack the grounding needed for real progress.
In other words, AI is more a mirror than a window, reflecting back what’s given instead of providing new views. Moving too quickly toward automation risks creating solutions that are accessible, but ultimately incomplete.
Supporting clinicians, not replacing them
We tend to fear AI in binary terms – it’ll either take our jobs, or we’ll resist. From 2014 to 2024, over half a million (516,790) lives were lost to suicide in the United States. We don’t have time for binary thinking, and we need to start considering ways that AI can work alongside mental health care workers, instead of how it might replace them.
Clinicians today manage heavy caseloads, increasing administrative demands, and fragmented patient information. Important details are typically spread across systems, obscuring the full patient picture. AI can help bring that picture into focus.
By analyzing larger data sets, AI can surface trends in symptoms and highlight changes over time, giving clinicians a clearer, more complete view of patient progress. It can support ongoing tracking, flag moments that may require closer attention, and take on routine tasks like documentation. With more of this work handled in the background, clinicians are able to spend more time focused on their patients and apply their judgment where it matters most.
Data as a differentiator
One of the challenges in mental health care has always been visibility. Progress is often observational, measured through conversation and periodic check-ins. While those metrics remain essential, they can make it difficult to track long-term change with precision. More consistent data changes that.
Regular assessments, simple surveys, and ongoing patient input can create a clearer view of how someone is doing between sessions. Over time, this builds a record of progress that goes beyond memory or single moments. The key is combining the qualitative power of human experiences with the quantitative power of machine data. For clinicians, this creates a stronger foundation for decision-making. For patients, it gives them a clearer sense of progress.
Where AI fits in mental health care
As AI becomes more embedded in mental health care, the conversation is not just about what it can do, but where it can become risky without the right boundaries in place.
That urgency is growing. Nearly 3 in 10 U.S. adults have already used a digital tool to support their mental health, and among those users, almost half report turning to general-purpose chatbots for guidance. These tools are often more accessible, more affordable, and for many, more comfortable than speaking with another person. But increased use does not always translate to appropriate use.
AI can be a helpful entry point for people navigating mild or early-stage concerns. It can offer structure, surface common coping techniques, and make support feel more accessible. But mental health needs rarely stay static, and not all conditions can be safely supported in that way.
As symptoms become more complex or persist without improvement, the limitations of AI become more pronounced. Mental health conditions are shaped by a combination of biology, personal history, relationships, and environment. That full context is difficult to capture through a chatbot, and even harder to interpret without clinical training .
In some cases, the risks are more direct. AI systems are designed to be responsive and affirming, but without clinical judgment, that can lead to reinforcing harmful or inaccurate beliefs. For individuals dealing with more serious conditions such as psychosis or severe mood disorders, that kind of feedback can be destabilizing rather than supportive .
The line, then, is not just about specific tasks like diagnosis or treatment planning. It is about severity, trajectory, and context. When symptoms begin to interfere with daily functioning, when progress stalls, or when conditions become more complex, care requires a level of interpretation and accountability that AI alone cannot provide. At that point, clinician involvement is not optional. It is essential.
This is where a more responsible model takes shape. AI can support care behind the scenes by helping clinicians track changes between visits, identify patterns, and surface insights that make treatment more proactive and informed. Used this way, it strengthens care without becoming the source of it .
As these tools continue to evolve, maintaining that boundary will be critical. Not to limit innovation, but to ensure it is applied in ways that are safe, effective, and grounded in the realities of patient care.
A responsible path forward
AI has the potential to improve access to care, reduce strain on clinicians, and facilitate better outcomes. Those benefits are real, and the industry has an opportunity to lead with clarity. That begins with transparent communication about how AI is used, what role it plays in care, and where human oversight remains central. Patients deserve to understand the systems that support them.
Mental health care does not need to become less human to become more effective. The most effective (and responsible) path forward is one built on partnership. AI provides insight, efficiency, and scale, while clinicians provide judgment, empathy, and compassion. Together, they create a system that can reach more people, while still maintaining the quality of care that patients desperately need.
Photo: metamorworks, Getty Images
Paul Kim is the founder and CEO of Sensible Care, an online behavioral health provider he launched in 2017 to expand access to mental healthcare. An Army veteran, Kim enlisted after 9/11 and served in Iraq, where he developed PTSD, an experience that shaped his perspective on gaps in care for service members and led him to start the company. He has served in multiple roles in the military, including as an infantry soldier and later as a signal officer, and previously worked in healthcare operations as a business manager at Oak Health Center. Kim holds a degree in philosophy from California State University, Los Angeles. Under his leadership, Sensible Care has grown to provide telepsychiatry and therapy services nationwide, with a focus on building human centered, outcomes oriented relationships between patients and providers and improving access for veterans and their families.
This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.
