Answers are grounded in this paper and cite their sources.
Introduction Artificial intelligence (AI) technologies now mediate romantic connections through two structurally distinct platform types: dating apps, where AI facilitates human-to-human connections, and AI companion platforms, where the AI itself becomes the object of the relationship. Their terms of service, privacy policies, and community guidelines operate as quasi-policy texts, defining what users may feel, disclose, and expect in a growing domain of intimate cultural practice and allocating responsibility for the emotional harms that practice produces. Methods This study conducts a comparative analysis of 49 governance documents from 14 platforms, drawing on Fairclough's three-dimensional model and the theoretical framework of emotional capitalism. Results and Discussion The analysis identifies two structurally distinct models of emotional extraction. The matching model of dating apps constructs emotion as a means to human connection, extracts behavioral signals at breadth through a standardized corporate template, and absolves responsibility for relationship outcomes between users. The bonding model of AI companions constructs emotion as the end itself, extracts affective histories at relational depth, and shirks responsibility for the emotional consequences of a relationship with the platform. Despite these structural differences, both models write AI-mediated emotional bonds into existence through conversational governance language while simultaneously writing the platform out of accountability. Conclusion A critical subfinding is the identification of an inverse governance pattern. The platforms that most aggressively solicit romantic and sexual bonding document the fewest emotional safety protections.
No takes yet. Share an insight, caveat, or question.
Yigang Liu (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: