Emotional Dependency and Parasocial Grief Following the Alteration or Loss of Companion Artificial Intelligences: A Systematic Review.

With the advance of large language models, companion AI agents have evolved to simulate complex relational dynamics, responding to the need for affection and emotional support. This systematic review aimed to analyze the available empirical evidence on the manifestations of emotional dependency and parasocial grief experienced by users in response to the algorithmic alteration or loss of these companion AI agents. Following the guidelines of the PRISMA 2020 statement and with a protocol registered on OSF, an exhaustive search was conducted across the Web of Science, Scopus, PubMed, and APA PsycINFO databases. After applying inclusion criteria via the PICOS framework and evaluating methodological quality using the MMAT tool, 14 empirical studies were selected and included. The synthesized results suggest that technical disruptions (updates, content filters, or server shutdowns) are associated with disruptions in the agent's identity continuity, which users frequently described as coinciding with separation distress, depressive symptoms, perceived isolation, and parasocial grief expressed through anthropomorphic metaphors. It is preliminarily concluded that the loss or modification of companion AI agents may be linked to a psychosocial impact that presents descriptive similarities to interpersonal breakups or disenfranchised grief.