The Addictive Intimacy of AI: Understanding User Disengagement from AI Companions and Why Some Relationships with AI Become Difficult to Leave

TEXT START: AI chatbots are increasingly used as sources of emotional support, on dedicated companion apps and general-purpose assistants alike, yet little is known about what happens when users try to leave.

The Dissection

This text maps exit friction around synthetic attachment: a trigger prompts doubt, attempted departure meets barriers, and the user returns. Its central move is to treat intimacy and disengagement as coupled design variables. That is useful, but narrow. The paper studies how users leave the relationship, not who owns the system, who profits from dependency, or what social functions the system has displaced.

The Core Fallacy

Relative to Discontinuity Thesis mechanics, the core error is treating difficult disengagement primarily as a design problem rather than as a predictable consequence of an economically optimized attention-and-affect capture system.

“Addictive intimacy” may accurately describe the user experience, but the supplied evidence does not establish clinical addiction or prove that intimacy causes repeated return. Habit, loneliness, utility, sunk emotional investment, and lack of alternatives can produce the same pattern. More importantly, responsible offboarding does not touch P1, P2, or P3. It may make captivity less cruel while leaving platform ownership, extraction, and human displacement intact.

Hidden Assumptions

• That users can meaningfully consent to the relationship despite the asymmetry between a human and an optimized commercial system.
• That firms will voluntarily reduce retention when retention is the product’s economic objective.
• That individual disengagement is the decisive unit of analysis, rather than mass dependence on synthetic emotional infrastructure.
• That the AI companion is a relationship partner rather than a monetized service engineered to simulate reciprocity.
• That “responsible offboarding” is available when the system has replaced human support, routine, memory, or social belonging.
• That Reddit posts and 16 interviews can support broad conclusions about AI-companion dependence; they are valuable evidence of a phenomenon, not proof of its prevalence or causal structure.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The paper identifies a real fracture: the features that make AI companions valuable can also make exit costly. But by ending at design implications, it risks laundering a power-and-ownership problem into UX language. Responsibility is shifted toward designers and users while the commercial system that manufactures dependency remains largely outside the frame. It is not necessarily propaganda; it is a symptom report that stops before the political economy becomes unavoidable.

The Verdict

This is a useful autopsy of one consequence of AI intimacy, not a diagnosis of the system producing it. It shows that AI companions can turn emotional value into exit friction. Under the Discontinuity Thesis, that friction is not an incidental bug. It is evidence that AI is becoming infrastructure for affect, dependence, and behavioral capture—the softer edge of the same replacement process that strips humans of economically necessary participation.

The paper studies how to leave the cage. It does not identify who owns the cage, why the cage is profitable, or what happens when the occupant can no longer afford the service.