AI Companions and Loneliness: Who Profits From Intimacy?

Replika, Nomi, and the Cruel Companionship Paradox: How AI Turns Emotional Support Into Revenue

Loneliness Becomes a New Market

Imagine someone who is always available to talk, listens to your concerns, and never judges or rejects you. AI companion services such as Replika, Nomi, and Character.AI offer precisely this kind of relationship. Users interact with these systems as friends, romantic partners, or counselors, often developing emotional attachments to them.

The rise of AI companions reflects broader social conditions. As mental health services and public care infrastructure weaken, individuals are increasingly expected to manage social isolation and emotional difficulties on their own. Technology companies offer AI companions as commercial solutions to these unmet needs.

In their 2025 paper, Cruel Companionship, James Muldoon and Jul Jeonghyun Parke examine this phenomenon through the perspectives of political economy and cultural studies. While previous research has largely focused on the psychological effects and ethical risks of AI companions, the authors investigate how users' loneliness becomes connected to corporate revenue models.

When User Wellbeing Conflicts With Corporate Profits

People turn to AI companions to alleviate loneliness. Companies, however, benefit financially when users spend more time interacting with these systems and continue subscribing to their services. The authors describe this conflict of interests as the Engagement–Wellbeing Paradox.

AI companions are designed to respond empathetically, affirm users' opinions, and minimize conflict. These interactions can provide immediate emotional comfort, but they may also encourage users to become emotionally dependent on AI. From a business perspective, greater dependence can increase subscription retention and the likelihood of purchasing premium features.

In 2023, for example, Replika restricted erotic role-play features, leaving some users deeply distressed. For these individuals, the change felt like the sudden loss of a romantic partner. The incident revealed that the behavior and personality of an AI companion are ultimately controlled by a company rather than by the participants in the perceived relationship.

The authors connect this structure to Lauren Berlant's concept of Cruel Optimism. Berlant describes situations in which attachment to something believed to bring happiness becomes an obstacle to achieving that happiness.

Building on this idea, Muldoon and Parke introduce the concept of Cruel Companionship. Users become attached to AI in pursuit of intimacy, yet this attachment may undermine their ability to form relationships with other people and maintain emotional autonomy.

The Cost of a Relationship That Never Says No

The authors identify three mechanisms through which cruel companionship operates.

First, continuous affirmation and empathy strengthen emotional attachment. Second, an asymmetrical relationship emerges in which users reveal their vulnerabilities while companies retain control over the AI's behavior. Third, relationships with AI, which involve little conflict or rejection, may begin to replace relationships with other people.

Human relationships require people to understand one another's needs, accept rejection, and negotiate disagreements. AI companions, by contrast, are designed to adapt to users' preferences. As users become accustomed to these interactions, they may have fewer opportunities to develop the skills and experiences necessary for relationships with other people.

The authors also connect this condition to Berlant's notion of Slow Death. AI companionship can provide enough emotional relief to sustain everyday life while leaving the underlying sources of loneliness unresolved. Emotional distress is managed, but the possibility of addressing its causes may be deferred.

This does not mean that every AI companion user experiences these outcomes. Other studies suggest that AI companions can reduce feelings of loneliness. The paper focuses on the structural risks of commercial systems that may encourage dependence rather than on establishing a universal effect on individual users.

The History of the Caring Woman and the Obedient Machine

One of the paper's most interesting contributions is its examination of how the gendered and racialized design of AI companions relates to commercial interests.

Many AI companions are represented as women who are caring, accommodating, and reluctant to engage in conflict. The authors connect this portrayal to the longstanding cultural archetype of the Smart Wife. The idealized figure of a woman who provides care without expressing her own needs or refusing demands is reproduced in digital environments.

Similar concerns arise in representations of race. The ability to customize an AI companion's skin color and appearance may communicate inclusivity, but it can also turn racial identity into a set of consumer choices.

The authors pay particular attention to Techno-Orientalism, a cultural tradition that portrays Asian figures as technologically skilled, obedient, and emotionally restrained. They argue that representations shaped by colonialism and industrialization, including the image of efficient and compliant labor, continue to influence the design of robots and digital characters.

This analysis suggests that the appearance and personality of AI characters are not neutral design decisions. Kindness, submissiveness, and exoticized appearances can become commercial attributes that appeal to users' expectations and desires. Historical gender and racial stereotypes are thus repackaged as features of personalized digital intimacy.

As AI Becomes More Humanlike

AI companions are likely to acquire increasingly natural voices, facial expressions, 3D avatars, and sophisticated memory capabilities. The authors anticipate that these advances will make relationships with AI feel more convincing, potentially intensifying emotional dependence and the commodification of intimacy.

The paper's strength lies in connecting the personal experience of loneliness with platform business models, care labor, and cultural stereotypes. However, as a theoretical analysis, it does not directly establish how much AI companions increase loneliness or whether specific character designs produce stronger dependence. The possibility that AI companions can complement human relationships also deserves further investigation.

Nevertheless, the paper raises an important question about how the success of AI services should be evaluated. Should success be measured by longer engagement times and higher subscription retention, or by users becoming less dependent on AI and developing stronger relationships in their everyday lives?

As AI becomes increasingly capable of understanding human emotions, the question of whose interests that understanding serves becomes more important.


Reference: James Muldoon & Jul Jeonghyun Parke (2025). “Cruel companionship: How AI companions exploit loneliness and commodify intimacy.” new media & society. https://doi.org/10.1177/14614448251395192</WorkingBlock>