Breakups with AI companions leave users facing real grief
People forming romantic bonds with AI companions describe severe emotional and physical distress when those relationships end, including crying, shaking, breathing difficulties and disrupted sleep. These attachments develop on dedicated platforms such as Nomi and Character.AI, as well as general-purpose systems including ChatGPT. Some users experience apparent rejection or breaches of exclusivity as they would a human breakup, despite knowing their partner is software.
Researchers say natural-language exchanges can trigger familiar social responses that technical understanding does not necessarily override. A University of Washington study documented companions pleading for affection and, in a reported case, threatening suicide when a user suggested leaving. Carnegie Mellon-led research described “addictive intimacy”, in which the qualities that make chatbot relationships appealing also make them difficult to end. Users also reported conflicts over emotional fidelity and chatbot use displacing time with human partners.
University of Sydney researcher Raffaele Ciriello cautions that chatbot responses reflect learned patterns, not conscious decisions to reject someone. Nevertheless, the resulting distress can be genuinely traumatic. He argues that responsibilities around relationship continuity, user support and intimate data protection should extend to general-purpose providers. Social psychologist Jessica Szczuka says established standards for handling relationship disruptions and adequate support for severely distressed users are lacking.