Emotional Reliance on Artificial Intelligence: A Scoping Review of AI Companionship and Mental Health Implications

Key Findings

• — 1,847 records screened across PubMed/MEDLINE, PsycINFO, Web of Science, Scopus and IEEE Xplore to March 2026; 46 studies met inclusion criteria.
• — The ELIZA effect is identified as the foundational mechanism of human-AI attachment, with documented cases of severe dependency including fatal outcomes.
• — Anthropomorphisation and artificial intimacy are attributed to adaptive, personalised design rather than to user naivety.
• — The authors propose Generative AI Dependency (GAID) as a construct mapped onto behavioural-addiction components, explicitly pending empirical validation.
• — Adolescents and lonely individuals are flagged as most vulnerable to forming exclusive affective bonds.
• — Named evidence gaps: longitudinal studies, validated screening tools and intervention protocols.

Methodology Notes

Scoping review following PRISMA-ScR with two independent reviewers and PCC-framework extraction; it maps evidence rather than pooling effects, so it supports claims about the shape of the literature and not about prevalence. The GAID construct is the authors' proposal and is unvalidated, which they state. MDPI venue with an unusually large and heterogeneous author list including a medical-imaging industry affiliation. Published 2026-09-02, CC BY 4.0, DOI 10.3390/psychiatryint7050194; metadata and full abstract verified at Crossref.

Sources