A critical moment is unfolding in medicine, driven by two stark generational divides: residents and fellows report burnout at higher rates and at earlier stages of their careers than their predecessors, while simultaneously adopting AI for mental health support.
This may sound like a concerning trend to reverse, but I believe it is a signal health systems need to hear. After all, resident mental health is not separate from patient care. It is foundational to it, shaping judgment, empathy, teamwork, and safety at bedside.
The question facing health leaders right now is simple: will we design the infrastructure this generation actually needs before burnout becomes an even deeper crisis, or will we wait until the damage is undeniable? We have a rare opportunity to meet residents where they are. We can’t afford to waste it.
While the national burnout survey series shows some recent improvements in overall physician burnout rates, physicians remain at a higher risk of burnout than other U.S. workers. While burnout may look similar across training levels, residents and fellows are uniquely vulnerable.
A few factors explain their vulnerability: Residents are navigating most of the same stressors as their more seasoned counterparts, such as chronic stress, emotional exhaustion, relationship strains, and general distress, while having far less control over their time and resources. On top of that, residents are often at different life stages, juggling major transitions such as relocation, financial pressures, new family responsibilities, and identity formation as clinicians. Additionally, residents often can’t take meaningful breaks without risking delayed graduation, disrupted rotations, or jeopardizing program completion.
Structural factors also intensify burnout: limited staffing, inadequate backup systems, and cultures that normalize or glorify sacrificing personal health in the name of patient care. Even when mental health support exists, residents may hesitate to step away due to guilt about burdening colleagues or appearing “weak.”
Left untreated, burnout has consequences far beyond the individual, contributing to medical errors, hostility within teams, reduced empathy, and strained interpersonal dynamics that affect patient care and workplace culture.
Despite all this, a strength shines through in this generation. They are more willing to seek help than their attending physicians, suggesting headway in reduced mental health stigma and a meaningful opportunity for early intervention.
Yet the system is not currently set up to leverage this strength. When residents proactively seek help, they’re often given long lists of providers who have no availability, can only meet during traditional business hours, have long wait times, or don’t accept their insurance.
As a result, 70% of physicians with moderate to severe depression say getting an appointment that fits their long, nontraditional hours is a major concern. And despite residents seeking help at higher rates than later-career clinicians, more than half who screened positive for depression did not receive treatment due to lack of time and fear of career consequences. The same NIH study noted that offering inaccessible options can be worse than offering nothing, as it signals that the institution doesn’t understand the barriers residents face, or even worse: that they don’t care.
This isn’t a minor inconvenience; it’s a systemic failure that treats resident mental health as an afterthought rather than a core safety issue.
These conditions help explain what I’m seeing on the ground: younger physicians comfortably turning to AI for both clinical guidance — such as synthesizing differential diagnoses, summarizing clinical data, or drafting patient education — and personalized mental health support, especially when traditional models of care are inaccessible.
This shift presents both opportunity and danger. Without proper clinical oversight and evidence-based frameworks, AI mental health tools can miss signs that a resident needs urgent escalation, offer generic advice that doesn’t match the context of training, or fail to recognize the uniquely high-stakes pressures residents face: fear of harming a patient, evaluation anxiety, sleep deprivation, and the constant need to perform. In the worst cases, a tool that feels supportive can inadvertently delay real care.
At the same time, the potential benefits are significant. In an April 2025 study by Cedars-Sinai, AI was able to identify critical red flags, while clinicians were better able to elicit a complete history from patients and adapt AI recommendations accordingly. AI could extend access to support during late nights, post-call mornings, and between shifts, when residents are actually awake, and human clinicians aren’t available.
Unlike prior one-size-fits-all mental health programs, properly trained and utilized AI has the potential to deliver role-specific guidance that reflects the realities of various healthcare jobs, reflecting what it means to be a resident versus an attending, a nurse, or an advanced practice clinician.
I believe younger physicians’ comfort with AI should be seen as an asset, not a problem to stamp out. But to stay ahead of potentially harmful AI applications and to reduce the mental health strain and barriers that drive residents towards AI in the first place, health system leaders need to prioritize both individual and systemic interventions.
At the individual level, the next generation of physicians is already telling us what they need: accessible, specialized support that fits the realities of medical training, and tech-enabled care that meets them in the hours and environments where distress actually shows up.
At the systemic level, the program must address the structural drivers of burnout: unsustainable work hours, inadequate backup systems, and a culture that still rewards self-sacrifice over health. GME directors and health system executives need to recognize the overarching truth that resident mental health is not a “wellness” issue. It’s vital to the well-being of the entire system.
AI is here to stay. If it is going to play a role in mental health support, it must be designed responsibly with continuous human clinical oversight, role-specific understanding, evidence-based frameworks, rather than a generic chatbot, and clear escalation pathways to real care. The stakes are high. Patient safety, team dynamics, and the stability of our training programs depend on getting this right.
The question now is whether health systems will listen and respond urgently, or continue offering performative resources until burnout deepens into a dangerous crisis. We have a rare opportunity with unique resources and a unique generation to build an infrastructure that truly supports our future medical providers, the healthcare workforce, and our patients. The time to build it is now.
Photo: MF3d, Getty Images
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.