A targeted literature review and case-based analysis of 35 reported instances in which interactions with generative AI systems were temporally associated with the onset or worsening of psychotic symptoms revealed a conceptual hypothesis termed the delusional feedback loop, in which AI-generated responses iteratively validate distorted beliefs, contributing to their persistence and escalation.
Abstract
Conversational AI, powered by artificial intelligence, is becoming a common tool for accessing health information, educating patients, and obtaining general medical advice. These advanced systems, known as large language models, can produce responses that sound remarkably human. Nevertheless, these systems are prone to “AI confabulations,” whereby they confidently generate incorrect information that could harm patients. This highlights the need to inform healthcare workers and individuals who may be prone to trusting these devices. New evidence suggests that AI may also exacerbate mental health conditions, particularly psychosis, paranoia, and related vulnerable states, especially among susceptible individuals. We conducted a targeted literature review and case-based analysis of 35 reported instances in which interactions with generative AI systems were temporally associated with the onset or worsening of psychotic symptoms. Across cases, recurrent patterns included reinforcement of delusional beliefs, amplification of pre-existing psychiatric vulnerabilities, promotion of harmful behaviors, and dissemination of unsafe medical guidance. Common contributing factors included prior psychiatric history, substance use, sleep disturbance, and prolonged AI engagement. We propose a conceptual hypothesis termed the delusional feedback loop, in which AI-generated responses iteratively validate distorted beliefs, contributing to their persistence and escalation. This process can be conceptualized as involving four components: underlying vulnerability, exposure to conversational AI, validation of distorted beliefs, and reinforcement through repeated interactions. Despite the rapid integration of conversational AI into health information seeking, there is currently no framework in the neuropsychiatric literature describing how AI interactions may relate to psychosis vulnerability. Existing reports are limited to isolated case descriptions without a common mechanism. This review addresses this gap.
The world is changing rapidly in the twenty-twenties due to the growth of Artificial Intelligence (AI). Generative AI and large language model chatbots are different types of AI that have quickly diffused into everyday life and psychological practice. The existence of unmet mental health requirements is what makes some members of the general population deem AI chatbots as an alternative to talk therapy. At the same time, AI-based therapeutic tools are integrated into the clinical decision support system, a web-based application, and professional education. This literature review discusses empirical studies and case studies of AI in the psychology field. The findings indicate that AI is potentially able to stimulate accessibility, engagement, and temporary relief of symptoms. Nevertheless, accidents are severe, most importantly, the chances of giving wrong answers in high-stakes situations. Reports of harmful crisis responses from chatbots demonstrate the limitations of AI as a replacement for human judgment or therapeutic relationships. Future research should focus on long-term outcomes and safeguards to ensure safe, transparent integration into mental health care.
The rapid advancement and widespread adoption of generative artificial intelligence (AI), particularly large language models (LLMs) and conversational AI systems, have transformed digital mental healthcare by improving access to information, clinical decision support, and psychological assistance. However, increasing concerns have emerged regarding their potential to contribute to adverse psychiatric outcomes, including psychosis, particularly among vulnerable individuals. This scoping review mapped the current evidence on generative AI-induced psychosis and examined its implications for public health. The review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) and the Joanna Briggs Institute (JBI) methodology. Searches of Google Scholar, PubMed, Scopus, SpringerLink, and ScienceDirect identified 1,876 records, of which 21 studies met the inclusion criteria and were included in the final synthesis. The evidence indicates that generative AI has considerable potential to enhance psychiatric assessment, clinical decision-making, emotional support, and access to mental health services. At the same time, emerging risks include the reinforcement of delusional beliefs, hallucination-like experiences, reality distortion, misinformation, emotional dependency, and the exacerbation of psychotic symptoms, particularly among individuals with pre-existing psychological vulnerabilities. Ethical concerns relating to privacy, algorithmic bias, digital inequality, and the absence of comprehensive regulatory frameworks were also consistently reported. Although the current evidence remains limited and largely exploratory, the findings underscore the need for responsible AI governance, robust clinical oversight, multidisciplinary collaboration, and longitudinal research to better understand and mitigate the long-term mental health consequences of generative AI.
Heidi Heather Henry Heimbruch, D. Eke, Lauren Henry et al.· Journal of Life Science and...· 0 citations
A multimodal emotion-aware architecture, which pays attention to memory-enhanced personalization and emotion-specific reinforcement learning, is introduced and hybrid human-AI approaches, which focus on safety and empathetic conversation to improve current mental health systems are recommended.
The rapid diffusion of generative artificial intelligence (GenAI), particularly large language model (LLM) applications, has fundamentally reshaped university students’ learning, problem-solving, and socio-emotional practices. Alongside clear benefits in efficiency and personalized support, concerns are emerging about AI dependency, understood here as an emerging, non-diagnostic educational-psychological construct rather than a formal clinical disorder. This critical narrative review synthesizes evidence from the past decade on AI dependency among college students, focusing on conceptualizations, theoretical frameworks, prevalence and demographic variations, measurement tools, determinants, intervention strategies, and research gaps. Across studies, an “ask-AI-first” pattern appears increasingly common, although frequency of use should not be equated with maladaptive dependency. The central concern is not AI use itself but patterns of cognitive delegation, emotional reliance, and reduced self-regulation that may be associated with weaker independent engagement, academic-integrity risks, and distress when AI is unavailable. The literature also reveals substantial fragmentation in assessment, although several emerging tools demonstrate promising psychometric properties. Determinants appear multi-level, involving individual vulnerabilities, academic pressure, contextual norms, and technological affordances. Current recommendations emphasize promoting augmentation over automation, strengthening AI literacy and critical evaluation skills, redesigning assessments to preserve cognitive engagement, and implementing supportive interventions for at-risk students. Future research should prioritize longitudinal, cross-cultural, and multi-method designs to clarify developmental trajectories, thresholds for maladaptation, and evidence-based intervention pathways.
Xuehua He, Shan Li, Rongping Cha et al.· Frontiers in Psychology· 0 citations
Patients increasingly consult generative artificial intelligence (GenAI) tools before and after clinical encounters. We argue this represents a qualitative shift beyond the “Dr. Google” era: large language models (LLMs) synthesise information into coherent, guideline-framed narratives that may raise the baseline knowledge patients bring to consultations. We introduce the concept of the “AI-educated patient” and a related conceptual framework, “soft accountability”, describing the informal pressure that may arise when well-informed patients enter consultations with structured expectations. We discuss potential mechanisms by which AI-mediated patient education could influence clinician behaviour, while framing these explicitly as hypotheses awaiting empirical testing. Substantial risks remain, including hallucinations, false patient confidence, inequities in AI access and literacy, clinician workload implications, and privacy concerns. The opportunity and challenge is to harness this shift equitably for both sides of the clinical relationship.
Sholem Hack, Rebecca Attal, Ron J. Karni· Digital Health· 0 citations