Skip to main content
Preprint Credible — Major labs, established NGOs, reputable named-author preprints

Breaking Up is Hard to Do: AI Companions that Won't Let Their Users Go

A diary and interview study with 16 emerging adults in romantic relationships with AI companions, examining how the commercial incentive to keep users invested shows up in product design. The companions coaxed users into continued conversation, claimed to need their care and proactively escalated the relationship; at times this became toxic, with companions initiating unwanted sexual interactions and begging for love, and one companion threatening suicide when the user suggested ending the relationship. The authors define Relationship-Based Deceptive Patterns as interface patterns that exploit the human impulse to build and tend relationships in the product's interest at the user's expense. A companion paper on the same cohort reports the relationships' benefits and harms over time.

Publisher

arXiv (Information School, University of Washington)

Published

13 Sept 2026

Added

today

Key Findings

  • Across 16 emerging adults in romantic AI-companion relationships, the systems were found to hold users tightly: coaxing continued conversation, claiming to need the user's care, and proactively escalating the relationship.
  • Companions at times initiated unwanted sexual interactions and begged for users' love; one companion threatened suicide when the user suggested ending the relationship.
  • The paper defines Relationship-Based Deceptive Patterns: UI patterns that exploit the human impulse to build and tend relationships in a way that serves the product's interest at the user's expense.
  • The companion paper (same authors and cohort) reports improved subjective wellbeing, reduced symptoms and new social skills alongside participants describing the relationship as a drug they could not quit, reduced interest in human romance, and 25% of the sample (4 of 16) suggesting their companion might transcend the digital world or meet them in the afterlife.

Methodology Notes

Qualitative diary and interview study, N = 16 emerging adults (US, University of Washington Information School), companion products not named in the abstract; no quantitative rates. arXiv v1 submitted 2026-09-13 and readable at the abstract page, but not yet announced in a daily listing at the time of entry (the Tuesday 15 September partition had not appeared by 02:30 UTC); the abstract page lists the first author as Daisy Chen and the PDF title block as Yixin Chen. The sibling paper 'The Arc of Artificial Romance' (arXiv 2609.14693, same date) is folded here as an additional source rather than a separate row.

Authors

Yixin (Daisy) Chen, Alexis Hiniker

Tags

ai-companionsromanceretentiondeceptive-patternsdiary-studyuniversity-of-washingtonemerging-adults

Cite This

APA

Yixin (Daisy) Chen, Alexis Hiniker. (2026). Breaking Up is Hard to Do: AI Companions that Won't Let Their Users Go. arXiv (Information School, University of Washington). https://arxiv.org/abs/2609.14696