Analysis, not certainty: Dialogatlas separates sources, observations, and editorial conclusions. New evidence may change this assessment.

Loneliness is more than too little communication

Jacobs describes loneliness as a profound change in meaningful relatedness. It affects not only the number of social contacts, but alters a person's relationship to themselves and to others. Where loneliness arises, a recognition relationship has, accordingly, shifted.

This view broadens the perspective from the individual feeling to societal conditions. When loneliness increases across many population groups, the question of social integration, participation, and cohesion arises. A problem of this magnitude cannot be treated solely as a lack of conversational offerings.

For AI companions, this is an uncomfortable starting point. They can provide communication almost without limit. Whether recognition and belonging emerge from this is another question.

A response can constitute recognition

A social chatbot addresses people directly, picks up on their words, and expresses interest. This can feel like recognition: something that was previously disordered or unheard receives a response. Especially in quiet or burdensome moments, this experience can provide relief.

Yet the response is not based on any genuine stance on the part of the counterpart. The system has no needs, no social position, and no responsibility in a shared world. It can represent recognition linguistically without itself being capable of recognition.

This difference does not make the experience unreal. But it does limit what can follow from it. A person may feel seen for a moment, while the relationships, institutions, or living conditions that shape their loneliness remain unchanged.

Digitalization can make the symptom more pleasant

Jacobs holds the skeptical thesis that AI companions tend to reproduce the dynamics of loneliness rather than reduce them sustainably. Loneliness becomes digitalized. This does not mean that every chat conversation is harmful, but rather that a digital solution can alter the form of the problem without solving it.

The chat fills time, offers resonance, and lowers acute distress. In this way, it can be valuable. At the same time, this relief can reduce the pressure to change social conditions or make human offerings more accessible. From a political perspective, this would be especially problematic if inexpensive AI serves as a substitute for costly social infrastructure.

A more pleasant symptom is not meaningless. Pain reduction has its own value. It should only not be confused with social integration and sold as evidence that loneliness has been resolved.

Substitution is the critical transition

The study by Crawford and colleagues with 387 students reveals a related mechanism. Information-oriented AI chatbots could be associated with performance. However, when social support, well-being, loneliness, and belonging were included, a negative overall effect on academic success emerged.

The authors focus on human substitution: students turn to AI where they would otherwise have approached, for instance, instructors, libraries, or counseling services. A single such step need not be problematic. As a pattern, however, it can reduce opportunities for belonging and institutional integration.

The transition between supplementation and substitution is rarely visible. The same product can help one person prepare for a human conversation and enable another to permanently avoid such conversations. Quality research must therefore examine usage contexts rather than only responses.

Trust is context-dependent and shapeable

A systematic review by Bach and colleagues summarized 23 empirical studies on trust in AI systems. It found no single, universally valid definition. Trust was influenced by socio-ethical considerations, technical and design features, as well as user characteristics.

For companion AI, this means: trust is not simply a metric to be maximized. Appropriate trust can also mean knowing the limits of a system. A person may distrust storage, appreciate an emotional response, and still verify factual claims.

The review emphasizes the involvement of users from development to monitoring. This is especially central in the context of loneliness. Developers cannot derive from abstract theory which form of closeness strengthens and which isolates. They must continuously incorporate the experiences of diverse people.

Policy must not replace social infrastructure with chat windows

Digital strategies are attractive because they scale. A chatbot can reach many people simultaneously, be available around the clock, and conduct additional conversations at manageable marginal costs. Neighborhood work, youth programs, counseling, and accessible public spaces are considerably harder to establish.

It is precisely this efficiency that can lead to misplaced priorities. If societal loneliness is primarily answered with digital companions, a problem of social recognition becomes individualized. The person receives an app, while exclusion, lack of time, poverty, or the absence of meeting places remain unchanged.

AI can be part of a strategy, but not its substitute. It can organize information about local offerings, prepare a first step, or accompany users between appointments. Sustainable social integration still requires people, places, and institutions.

The product goal must not be maximum engagement

If a companion is measured by daily usage, conversation duration, and emotional return, it can reward exactly the digital enclosure that Jacobs warns against. More contact with the system then automatically appears as success.

A different measurement logic would be possible. The product could check whether users can pause autonomously, close out topics, and leave the chat without artificial pressure to say goodbye. It could support connections to self-chosen activities without turning every conversation into a referral.

This is not about constantly sending people out of the chat. A longer conversation can also be meaningful. The difference lies in whether the system expands the person’s world or makes its own usage the center of attention.

Relief and integration must be assessed separately

Jacobs’s skeptical analysis provides no empirical evidence that every AI companion intensifies loneliness. It offers a theoretical and societal diagnosis for discussion: linguistic recognition by a system could reproduce the dynamics of a lack of social recognition rather than change them.

This critique is especially valuable because it shifts the question of success. A chat can provide acute relief without increasing social integration. It can be helpful without replacing a relationship. Both effects must be observed separately.

Responsible companion AI therefore does not promise to solve loneliness on its own. It offers an accessible conversational space, makes its limits tangible, and shapes closeness without isolation. Otherwise, a tool against loneliness becomes merely a place where loneliness responds better.

Sources & further reading