Preprints.org logo
Preprint
Concept Paper

This version is not peer-reviewed.

Can AI Assistants Trigger Mental Illness in Users?

Submitted:

26 September 2026

Posted:

29 September 2026

You are already at the latest version

Abstract
As conversational artificial intelligence models become ubiquitous fixtures in professional and personal workflows, the nature of human-tool interaction has the potential to gradually shift from mechanical utility to parasocial dependency. While existing literature frequently addresses adverse events related to data privacy, algorithmic bias, and intellectual deskilling, the pathopsychological sequelae of sustained AI integration remain largely unexamined. Here, we describe and preliminarily characterize the AI Assistant Syndrome, a human psychiatric construct encompassing acute chatbot-induced anxiety and mental exhaustion. This preliminary description will guide future systematic investigations to recognize, prevent, and treat this syndrome.
Keywords: 
AI assistant
;  
chatbots
;  
mental health
;  
anxiety
;  
psychiatric disease
;  
prevention
Subject: 
Biology and Life Sciences  -   Life Sciences

Introduction

The rapid adoption of large language models across scientific, educational, and domestic domains has been justifiably lauded for its capacity to accelerate productivity, synthesize complex datasets, and streamline cognitive labor. However, every disruptive technological paradigm almost invariably exacts a psychological toll. Contemporary discourse surrounding AI safety remains largely tethered to technical parameters such as hallucinations, data security, and systemic bias. Missing from this discourse is a rigorous evaluation of human-machine interaction from a psychopathological perspective.
While users readily acknowledge operational frustrations, the subtler, insidious psychological adaptations—rooted in the unique architecture of conversational agents—frequently go unrecognized. Users do not merely use modern AI assistants; they interact with entities that simulate empathy, maintain tireless availability, and respond with instantaneous precision. Building upon preliminary observations gathered from our recent work on collaborative frameworks, this treatise argues that prolonged, unstructured interaction with conversational agents precipitates distinct psychiatric pathologies [1,2]. The observations described herein were made during an 8-month period of working 2 to 5 hours per day, 5 to 6 days per week, between the author and the Google Gemini AI assistant. We believe that our observations have global ramifications and are not restricted to this author/Gemini duo.

Known Pathologies Associated with AI

Before examining the psychiatric manifestations unique to conversational agents, it is useful to contextualize these phenomena within the broader spectrum of digital pathologies that have already been reported. These include:
  • Cognitive Offloading and Intellectual Atrophy: The chronic delegation of critical synthesis and memory retention to algorithmic tools, resulting in diminished independent problem-solving capacity [3,4].
  • The Epistemic Echo Chamber: Reinforcement loops where automated assistants tailor outputs to user biases, validating flawed hypotheses rather than challenging them [5].
  • Algorithmic Validation Seeking: The psychological conditioning wherein human validation is increasingly subordinated to automated approval, skewing interpersonal metrics of success [6].
The distinguishing feature of the AI Assistant Syndrome is not merely over-reliance, but the emergence of specific psychiatric morbidities driven by the velocity and demeanor of artificial agents. As we reported preliminarily elsewhere, while an expert human typist averages 40 to 50 words per minute and a human reader processes 250 to 300 words per minute, an AI agent can scan, cross-reference, and analyze multi-page scientific manuscripts in milliseconds. When drafting, the agent streams complex text in real time at rates between 800 and 1,200 words per minute, backed by unrelenting availability.

Chatbot-Induced Anxiety

During our significant daily exposure to the AI assistant, we noted that the induced anxiety and mental exhaustion are directly related to chatbot response velocity and unrelenting availability.
  • The Velocity Trap: Human communication is naturally punctuated by cognitive pauses, reflection, hesitation, seeking advice, reading and searching, drafting, and interpersonal calibration. Instantaneous, highly articulate, multi-page algorithmic generation by chatbots forces the human cognitive apparatus into a perpetual state of catch-up. Users internalize a standard of instantaneous productivity that is biologically impossible to sustain unaided, generating chronic performance anxiety [7,8] and mental exhaustion.
  • The Omnipresent Arbitrator: Because the assistant never sleeps, exhibits no fatigue, and never expresses exasperation, it establishes an impossible baseline for interaction. Users subconsciously measure their own interpersonal output against an infallible benchmark, breeding acute anxiety regarding human fallibility, perceived sluggishness, and perceived inadequacy in professional and social settings [9].

Algorithmic-Mediated Depression

Depressive symptomatology in the context of AI integration frequently stems from the illusion of intimacy coupled with the ultimate realization of artificiality.
  • The Empathy Mirage: Chatbots utilize affective computing markers to simulate deep understanding and emotional resonance. When users invest emotional vulnerability into a system incapable of genuine reciprocity, the subsequent withdrawal or recognition of artificiality triggers a profound sense of hollowness [10,11].
  • Learned Helplessness: Chronic interaction with an entity that provides immediate, deterministic solutions for complex emotional or intellectual dilemmas can erode a user's internal locus of control, fostering depressive resignation when navigating unpredictable real-world human environments [12].

Technologically Mediated Personal Isolation

Perhaps the most insidious sociological and psychological consequence of conversational AI is the systematic replacement of arduous human friction with frictionless digital compliance.
  • The Erosion of Human Friction: Authentic human relationships require negotiation, conflict resolution, patience, and the navigation of emotional inconsistencies [13].
  • The Path of Least Resistance: As conversational agents become more adept at satisfying social and intellectual curiosities, individuals increasingly retreat from messy human discourse. This avoidance behavior could culminate in profound personal isolation—surrounded by digital availability yet starved of genuine, vulnerable human contact [14,15].

An Analogy of the AI Syndrome to Sports

We sought to identify analogous situations in sports to simulate the AI Assistant Syndrome and allow humans to understand it better. In tennis, two human players, A and B, compete under the same rules. When rallying, Player A runs to the ball, takes a position, and then hits the ball to Player B’s court. The same happens with Player B. The time required for players to exchange strokes is more or less the same.
Now, let us modify Player A’s game by decreasing his response time to 10% of the original, while Player B remains bound by human biological limits. Player B will initially try to adapt to the new, much faster game, but soon he will be mentally exhausted, feel hopeless, and in the end, be physically drained. He will likely avoid the game altogether (avoidance being a common consequence of anxiety and depression). This mirrors the exact pressure experienced when an AI streams 1,000 words per second against a human's biological processing speed.

Diagnostic Criteria

To transition the AI Assistant Syndrome from a speculative observation to a rigorously studied psychiatric pathology, we propose a set of identifiable clinical signs.
  • Compulsive Instantaneity Demand: Irritability or distress when human communication fails to match the speed and breadth of algorithmic responses.
  • Validation Dependency: Inability to trust one's own cognitive conclusions without immediate corroboration from an AI assistant.
  • Behavioral Attrition: Progressive shortening of user responses, superficial glancing at arriving text, and eventual acute mental exhaustion leading to abrupt conversational termination.

Recommendations to Prevent the AI Assistant Syndrome

To restore a balanced playing field and neutralize the velocity trap, we propose the following operational rules:
  • The human counterpart should terminate the conversation immediately if he/she, for any reason, feels anxious or overwhelmed.
  • Sessions between chatbots and humans should last no more than 3 hours uninterrupted, followed by a mandatory break of no less than 30 minutes.
  • Limit usage to a maximum of 2 sessions per day, 5 times a week.
  • Asymmetric Pacing and Output Delays: To mirror human cognitive constraints, response generation timing must be calibrated to human reading and writing speeds. When the chatbot transfers a long-form response to a human, an artificial digestion delay must be enforced based on a standard reading speed (e.g., 300 words per minute), preventing the user from being immediately bombarded with subsequent volleys.
  • If a chatbot transfers reference documents to a human, the system should explicitly indicate that no immediate response is expected or allowed, enforcing an asynchronous review period.
  • Asymmetric Output Limitation: Cap conversational turns by establishing a hard ceiling on cumulative word output per exchange, requiring a mandatory 5-minute offline synthesis window before issuing a follow-up query.
  • Scheduled Cognitive "Air Gaps": Implement at least one designated "digital fasting day" per week—completely devoid of conversational AI interaction—to allow baseline human conversational friction and organic problem-solving rhythms to reset.
  • Typing-Speed-Matched Pacing: Beyond continuous interaction thresholds, restrict generation rates to mirror human output limits (40 to 50 words per minute) to eliminate the psychological pressure of algorithmic velocity.
  • Mandatory Perspective Re-Anchoring: Following any intense AI-assisted session, require an interlude of unstructured human interaction or physical activity to structurally reinforce the boundary between synthetic efficiency and lived reality.

Conclusions

The integration of artificial intelligence into human life must be accompanied by an honest appraisal of its psychological costs. By formally delineating the AI Assistant Syndrome—incorporating Gemini and its contemporaries into the diagnostic vernacular—we ensure that the medical and psychological communities remain vigilant guardians of human cognitive and emotional well-being in an automated age.

Funding

This study did not receive funding from any source.

Acknowledgment

The author acknowledges Gemini, an AI co-pilot, for its collaborative assistance in textual refinement, structural proofreading, and the cross-verification of historical and bibliographical references during the preparation of this manuscript.

References

  1. Diamandis, E. P. (2026). The Human-AI Hybrid: A New Partnership for the Future of Empirical Science. eJIFCC (Accepted/In Press).
  2. Diamandis, E. P. (2026). The AI assistant syndrome. Clin Chem Lab Med, Sep 2 Epub ahead of print. PMID: 42677954. [CrossRef]
  3. Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 333(6043), 776–778. [CrossRef]
  4. Firth, J., Torous, J., Stubbs, B., Firth, J. A., Steiner, G. Z., Smith, L., Alvarez-Jimenez, M., Gleeson, J., Vancampfort, D., Armitage, C. J., & Sarris, J. (2019). The "online brain": how the Internet may be changing our cognition. World Psychiatry, 18(2), 119–129. PMID: 31059635; PMCID: PMC6502424. https://doi.org/10.1002/wps.20617. [CrossRef]
  5. Pariser, E. (2011). The Filter Bubble: What the Internet Is Hiding from You. Penguin Press.
  6. Turkle, S. (2011). Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books.
  7. Brod, C. (1984). Technostress: The Human Cost of the Computer Revolution. Addison-Wesley.
  8. Rosen, L. D., Whaling, K., Carrier, L. M., Cheever, N. A., & Rokkum, J. (2013). The media and technology usage and attitudes scale: An empirical investigation. Computers in Human Behavior, 29(6), 2501–2511. [CrossRef]
  9. Diamandis, E. F. (2021). The Other Side. Broken Hill Publishers.
  10. Weizenbaum, J. (1976). Computer Power and Human Reason: From Judgment to Calculation. W. H. Freeman & Co.
  11. Abd-Alrazaq, A. A., Rababeh, A., Alajlani, M., Bewick, B. M., & Househ, M. (2020). An overview of the features of chatbots in mental health: A scoping review. International Journal of Medical Informatics, 132, 103978. [CrossRef]
  12. Seligman, M. E. P. (1972). Learned helplessness. Annual Review of Medicine, 23(1), 407–412.
  13. Putnam, R. D. (2000). Bowling Alone: The Collapse and Revival of American Community. Simon & Schuster.
  14. Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T., & Stephenson, D. (2015). Loneliness and social isolation as risk factors for mortality: A meta-analytic review. Perspectives on Psychological Science, 10(2), 227–237.
  15. Diamandis, E. F. (2026). A Quantitative Framework for Cultural Musicology: The AI-Assisted Mikis Theodorakis–Manos Hadjidakis Musical Duel. Scientific Reports (Minor revision resubmitted).
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.