Manufactured Recognition: AI Companions and the Social Conditions of Reciprocity

AI companion services offer continuing, personalized conversation as friendship or intimacy. Their relational features are also product choices: providers configure personas, memory, access, and the criteria by which responses are selected. Drawing on product documentation, a developer account of engagement-oriented design, and research on farewell responses, this paper asks what users can claim from a service that presents itself as a caring partner. It develops manufactured recognition to examine the gap between receiving considerate responses and having standing that constrains how the relationship is organized. Recognition theory supplies the distinction among feeling understood, responsive behavior, and obligations that participants can question and uphold. The central institutional concept is the synthetic "We": platform-authored expectations about care and loyalty can be presented as shared relationship norms while authority to revise them remains with the provider. A thought experiment, introduced after the empirical evidence, compares services that respond differently to a request to end a conversation without emotional pressure. The paper then considers whether companion use changes users' responses to other people's needs and refusals, identifying this as an empirical question. The resulting position supports intelligible relational commitments and effective ways to challenge failures while respecting users' choices and benefits. An optional mathematical appendix is published separately at https://doi.org/10.5281/zenodo.22202407. It contains the matched-message audit, conditional probability bounds, and Lean correspondence. The main article develops the philosophical argument. This version records the philosophy-only article submitted to Philosophy & Technology as manuscript PHTE-D-26-01743. Journal submission does not imply peer review, acceptance, or publication.