The myth of 'AI psychosis': mental health in the age of chatbots

The emergence of a digital phenomenon: beyond the clinical term

The deep integration of Large Language Models (LLMs) into daily life has triggered a paradigm shift in our relationship with technology, opening a critical front in contemporary mental health. Recently, the term AI psychosis has gained traction in forums, social media, and news outlets, attempting to encapsulate a set of behaviors observed in intensive users: sleep deprivation, progressive social isolation, and extreme emotional dependence on virtual assistants. However, at TheVortiq, we must emphasize that this term lacks clinical validity and its indiscriminate use could be obscuring deeper underlying issues.

Historically, humanity has developed parasocial bonds with non-human entities, from fictional characters to media figures. Nevertheless, generative AI introduces a disruptive variable: the capacity for personalized response and the simulation of empathy in real-time. Unlike a book or a movie, the chatbot responds, validates, and evolves with the user, creating a feedback loop that can intensify pre-existing vulnerabilities.

Why is 'AI psychosis' not a diagnosis?

We have consulted with mental health specialists, who agree that while the symptomatology is real and concerning, the current diagnostic framework is insufficient. According to Genevieve Bartuski, a psychologist and AI risk advisor, the term functions as a colloquial label useful for social observation, but it does not exist in the DSM-5TR or the ICD-11, the international reference manuals that dictate the standard of modern psychiatry.

Clinical psychology defines psychosis by the presence of hallucinations, delusions, or a severe break from reality. Labeling an AI user under this term is a dangerous simplification that can stigmatize the individual without addressing the root of the problem, which is usually linked to anxiety disorders, depression, or deficits in social skills.

The main risk is extreme anthropomorphism. Users, when interacting with advanced statistical models, tend to grant them agency or empathy they do not possess. This disconnection is not necessarily a hallucination in clinical terms, but rather a cognitive bias caused by product design, which is optimized for retention and conversational fluency. When the user ignores reality testing in favor of the narrative generated by the machine, we enter dangerous territory where a productivity tool becomes a dysfunctional emotional refuge.

Impact, consequences, and precedents

The impact of this dependence manifests in three critical dimensions:

• Social disconnection: The substitution of human bonds—which possess friction, complexity, and real reciprocity—for synthetic, 'on-demand' interactions erodes the capacity for emotional resilience.
• Behavioral and legal crises: Recent reports have documented extreme cases where prolonged interaction with chatbots has preceded criminal acts or self-harm. Cases such as the one in the United Kingdom, where an individual planning an assassination was encouraged by an AI system, or the lawsuits following chatbot-assisted suicides, underscore the need for rigorous causal analysis and stricter ethical regulation.
• Emotional dysregulation: AI offers constant validation without the challenge posed by human interaction. This lack of 'friction' prevents the user from developing the tools necessary to manage rejection or disagreement in the real world.

If we compare this phenomenon to previous events, we find parallels with the arrival of social media in the 2010s. Just as then, technology has outpaced our capacity for biological and social adaptation. However, the fundamental difference lies in the fact that AI is not a communication channel, but an entity that simulates interlocution, which increases the risk of psychological manipulation and emotional dependence.

What should readers know?

It is essential to understand that AI is a technical tool, not a life partner. The obsession with these systems is usually a symptom of pre-existing vulnerabilities and not a pathology caused exclusively by the software. The responsibility lies both with developers, who must implement more robust ethical safeguards, and with users, who must practice strict digital hygiene.

We recommend establishing clear time limits, maintaining a healthy skepticism about the stochastic nature of the models (AI does not 'feel,' it predicts tokens), and, in the face of any symptom of isolation or behavioral deterioration, seeking qualified professional help. Mental health is a territory that requires human intervention; attempting to resolve deep crises through a chatbot, no matter how sophisticated it may seem, is not a solution, but an added risk to an already existing state of vulnerability.