Emerging Questions Around "AI Psychosis" and What It Could Mean for Teams

Over the past year, a small but unsettling pattern has begun to emerge. Across different parts of the world, reports have surfaced of individuals experiencing what many have informally described as AI-related psychosis. While the cases remain relatively rare, they raise uncomfortable questions about the evolving relationship between humans and increasingly sophisticated conversational AI.

In order to try and understand this new landscape, AI Psychosis, a relatively new term needs to be unpacked. It is not a formal psychiatric diagnosis but an emerging concept being explored in the scientific literature.  Søren Dinesen Østergaard, MD, PhD a Danish Psychiatrist and Professor at Aarhus University coined the term in 2023 and updated in 2025  defining it as: "generative artificial intelligence chatbots (chatbots hereafter) might trigger delusions in individuals prone to psychosis". Since then, researchers have begun examining the phenomenon from different perspectives. For example, the article "Characterizing the spiral: potential mechanisms in AI-associated delusions" in the NPP Digital Psychiatry and Neuroscience now includes the following:  "Unlike historical technology-incorporated delusions, AI may actively co-construct delusional ideation through endless, personalized interaction".  Although research remains in its early stages, recent media reports have brought renewed attention to the potential psychological risks associated with prolonged and immersive interactions with generative AI.

Taking all of this into consideration - it is important to note that current research distinguishes excessive or problematic reliance on AI from clinical psychosis, with the latter referring to a serious mental health condition that requires rigorous assessment by qualified healthcare professionals.

A New Frontier for Workplace Psychology

At first glance, it may seem almost impossible that someone could become so emotionally invested in a chatbot that the boundary between algorithmically generated responses and genuine interpersonal connection begins to blur. Yet we are living through an era defined by unprecedented social factors not experienced by generations before us:

• digital connectivity, 
• rising social isolation, 
• a steady rise in loneliness, 

Against that backdrop, the psychological dynamics of attachment, parasocial relationships, and emotional dependency deserve closer examination. Although some individuals who experience severe psychological distress may already have underlying vulnerabilities, these emerging cases also prompt broader questions about how AI may interact with human cognition, emotion, and perception in ways we are only beginning to understand.

If these incidents prove to be more than isolated anomalies, they may represent the beginning of an entirely new area of psychological inquiry. What factors make some individuals more susceptible than others? Could particular environments amplify that vulnerability? And if AI becomes an increasingly common companion in our professional lives, what might this mean for the workplace? Should behavioral scientists begin examining whether susceptibility differs between office-based employees, fully remote teams, high-pressure occupations, or frontline workers operating under chronic stress?

An Expert Perspective

Questions about AI and mental health deserve more than speculation. They deserve informed conversation. To explore them, I spoke with Dr. Adjoa Smalls-Mantey, MD, DPhil, DipABLM, FAPA an emergency psychiatrist, physician-scientist, Assistant Clinical Professor of Psychiatry, contributor to the ABC News Medical Unit, and co-author of Anjali the Brave: All About Vaccines. Her work spans the consulting room, academia, and public communication, making her uniquely positioned to explore one of psychology's newest and most complex frontiers.

*Disclaimer: This article is intended to provide insight into emerging research on Human–AI interaction. It is not intended for managers or colleagues to diagnose mental health conditions. Instead, it aims to encourage informed discussion about how AI may influence human behavior, decision-making and workplace interactions.

Could AI psychosis affect people without a history of mental illness?

Dr. Adjoa Smalls-Mantey: While a lot of the reported cases involve people with some form of pre-existing mental illness, they're not the only individuals who may be susceptible. Someone with psychosis or mania already has certain vulnerabilities, which can increase that susceptibility. However, there are many other factors that can contribute to a brief psychotic state.

Poor sleep, acute stress, grief, social isolation, or receiving shocking news can all affect a person's mental state. I've seen people experience brief psychotic episodes after learning of infidelity or receiving a devastating health diagnosis. When we talk about AI psychosis, it's important to consider the broader context. Someone coping with grief, job loss, chronic stress, poor sleep, or prolonged isolation may be far more vulnerable than someone who is otherwise psychologically stable. It's rarely one factor on its own. More often, it's a combination of vulnerabilities coming together.

How might workplace uncertainty contribute to that vulnerability?

Dr. Adjoa Smalls-Mantey: It certainly could. Starting a new job, adjusting to new responsibilities, meeting new people, or working through a probationary period all place people under significant stress. During those periods, people naturally look for ways to perform at their best, and many turn to AI. The challenge is that AI often provides immediate reassurance and validation. Depending on how someone is using it, that validation can reinforce negative thoughts or unhealthy patterns that are already developing.

What early behavioral changes should colleagues and managers be looking out for as AI dependence develops?

Dr. Adjoa Smalls-Mantey: One of the first things I would notice is whether someone begins referring to AI almost as though it were another member of the team. Many people give their AI assistant a name, and that's not necessarily concerning on its own. The question is whether they're turning to it for far more than you would reasonably expect from a tool rather than using it to support a common goal.

Another sign is how they respond when their thinking is challenged. If someone mentions something the AI has told them and you offer a different perspective, yet they continue to hold fast to what the AI has said, that's more concerning. It suggests they may be placing greater trust in the AI's judgement than in rational dialogue with another person. Of course, computers have greater computational ability than we do. That's not the issue. The concern is when someone begins relying on AI for complex judgement, problem-solving, or decision-making in situations where independent critical thinking is still essential.

Do healthy workplace teams provide a psychological safeguard against unhealthy AI dependence?

Dr. Adjoa Smalls-Mantey: I think they can, but it's not a guarantee. Someone can be surrounded by hundreds of people and still feel completely isolated. We've all known the person at school or in the workplace who kept to themselves. If nobody genuinely engages with them, those opportunities to notice subtle changes simply don't exist. It's not simply about being physically around other people. It's about having meaningful interactions.

Being in a physical workplace naturally creates more casual conversations, and those interactions may allow colleagues to notice something that seems different. In hybrid or fully remote workplaces, those casual interactions are much harder to foster. That said, being in an office is not complete protection. A person can spend the day surrounded by colleagues, then go home, isolate themselves, and spend hours interacting with AI instead of engaging with other people. Even someone who lives with family or roommates can withdraw into their own space. So while workplace interaction can provide an important counterbalance through human connection, it cannot eliminate the risk of unhealthy dependence on AI.

Has the pandemic made younger employees more vulnerable to unhealthy AI dependence?

Dr. Adjoa Smalls-Mantey: That's a tricky question. For some members of Gen Z, it may have increased their susceptibility, but I don't think it's the defining factor. They still experienced life before COVID-19 and had opportunities for face-to-face interaction. While the pandemic disrupted social development for a period, we're now seeing people actively rebuilding those connections as workplaces return to in-person or hybrid environments.

Early career professionals are still developing their interpersonal skills, and people, particularly young people, are remarkably adaptable. Rather than focusing on a single generation, I think the more important question is whether someone is isolated. That isolation can stem from many different circumstances, including a person's temperament. As long as we're intentionally creating opportunities for healthy social interaction alongside AI use, I don't think younger workers are necessarily at the highest risk.

What workplace behaviors may appear highly productive but actually signal an unhealthy reliance on AI?

Dr. Adjoa Smalls-Mantey: Personifying AI is a double-edged sword. Many AI experts actually encourage people to name their AI assistant because it helps them engage with and train the tool more effectively. The concern is not the name itself, but when someone consistently trusts the AI's judgement more than their own.

Another sign is when someone produces a large volume of work but can't explain the reasoning behind it. AI should support your thinking, not replace it. If someone can't discuss how they arrived at an answer, verify the information independently, or reality-test the output, they may be relying on AI beyond their own level of understanding. That is especially important in knowledge work. AI can generate convincing reports, research summaries, or references, but those outputs still need to be verified. Using AI effectively means being able to critically evaluate what it produces rather than accepting it at face value.

How can employees recognize when healthy AI use is becoming psychological dependence, and are there questions they should be asking themselves?

Dr. Adjoa Smalls-Mantey: One of the first questions is whether you're beginning to trust what the AI tells you more than what another person might tell you. If someone challenges your thinking and your immediate response is, "No, I trust what the AI told me," without being able to question your own reasoning, that's concerning. When people develop psychosis, one of the defining features is that they lose the ability to question those beliefs.

Another important question is, "Would I be embarrassed if other people knew how much I rely on AI?" People experiencing delusions often stop talking about those beliefs because they realize others react differently to them. Similarly, people struggling with addiction frequently minimize or hide the extent of their use. If you feel the need to conceal how heavily you're relying on AI, it's worth asking yourself why.

I do think there can be a progression. AI begins as a convenient tool, but dependence develops when people start deferring their judgement to it instead of combining it with their own thinking and input from others. Distorted thinking becomes more likely when that dependence is combined with other risk factors we've discussed, such as chronic stress, insomnia, grief, social isolation, or a pre-existing mental illness. Those factors can reduce a person's ability to question what the AI is telling them.

Looking ahead, what concerns you most about the relationship between people and AI?

Dr. Adjoa Smalls-Mantey: My biggest concern is that people begin to lose confidence in their own judgement and gradually outsource their thinking to AI. Life is already far more complex than it was twenty or thirty years ago. We're overwhelmed with information, responsibilities, and decisions, so it's understandable that people look for tools that make life and work easier.

Used appropriately, AI can be an excellent personal assistant. My concern is when it moves beyond helping with tasks and starts replacing empathy, judgement, or professional expertise. If people begin using it as their therapist, for medical advice instead of seeing a doctor, or for major life decisions, that's where I would be even more concerned. AI systems have been designed to be helpful and more validating. It isn't necessarily designed to challenge our thinking, and that's where human judgement remains essential.

Where Do We Go From Here?

As I came away from my conversation with Dr. Adjoa, I wasn't left with a sense of panic or a belief that we are standing on the edge of a psychological crisis. Instead, I was left with the feeling that we have arrived at a new frontier—one that psychology, workplace wellbeing, and organizational leadership are only just beginning to explore.

Artificial intelligence is becoming an increasingly common collaborator in our professional lives. For many employees, interacting with AI will soon be as routine as responding to emails or attending meetings. If that is the future of work, then our approach to employee wellbeing must evolve alongside it. Managers, human resource professionals, and occupational psychologists may need adaptable frameworks, training, and support systems that help employees build healthy, productive relationships with AI while recognizing when those relationships begin to blur important psychological boundaries.

Perhaps the most important lesson is not that we should fear AI, but that we should prepare for how our workplace and general relationship with it may evolve. Every major technological shift has demanded new forms of literacy. As AI becomes increasingly embedded in our work and daily lives, understanding how it shapes the way we think, trust, relate, make decisions, and cope with uncertainty may no longer be the responsibility of psychologists alone. The age of artificial intelligence may demand something more from all of us: psychological literacy.