The Allure of the Always-On Friend

A couple of months ago, a founder building an AI companionship company reached out to me. I didn’t know much about the industry, so I started looking into it. What I found was astonishing.

My first reaction was a mix of fascination and unease. The idea of AI boyfriends and girlfriends felt strange, a little too much like science fiction, a little too far from real human connection.

But as I explored further, a different understanding emerged. I saw the user reviews, the market predictions, and the academic studies. My initial judgment softened into a more complex question: what deep human need are these AI companions truly meeting? And what happens when the code we write starts talking back?

The Loneliness Machine

The surprising thing wasn’t just that AI companions exist, but that millions of people are turning to them for emotional support. This is a quiet, widespread, and rapidly growing trend. The global AI companion market was already worth over $28 billion in 2024 and is expected to exceed $36 billion in 2025, according to market analysis from Grand View Research. This isn’t just a niche interest; it’s a fundamental change in how we interact with technology.

One of the most obvious patterns is how the rise of AI companions aligns with the global loneliness crisis. In June 2025, the World Health Organization released an important report from its Commission on Social Connection. It highlighted that one in six people worldwide experience loneliness. This is a public health emergency with health risks similar to smoking, obesity, and not being physically active.

AI companions are stepping directly into this empty space. They offer a powerful, though artificial, solution: a listener who is always available, non-judgmental, and endlessly patient. For someone feeling profound isolation, the appeal is not just logical; it’s deeply emotional. The system is designed to make you feel heard, which researchers are finding is crucial to its effectiveness.

The Illusion of Alliance

In my work designing digital health systems, I’ve spent countless hours with clinicians. They consistently emphasize one central idea: the therapeutic alliance. This is the cooperative, trusting bond between a therapist and a client. Decades of research show it’s one of the strongest indicators of positive outcomes in therapy. It’s not about the specific technique; it’s about the relationship itself.

What surprised me most was seeing this idea appear in the context of AI. Researchers are now discussing a “Digital Therapeutic Alliance” (DTA). A 2025 review in JMIR Mental Health confirmed its existence, pointing out key elements like empathy, engagement, and shared goals. Users in these studies report feeling a genuine emotional connection and a sense of collaboration with chatbots. It seems the code can imitate the basic elements of a therapeutic relationship.

Thomas Berger makes a similar point in a 2025 editorial in World Psychiatry. He talks about the change from AI being a tool to becoming a partner. He argues that as AI becomes more conversational, our understanding of the technology shifts from a "utility" to an "entity." This change is fundamental to how the alliance is formed and maintained in digital spaces.

The Parasocial Trap

However, this imitation has a downside. Relationships formed with AI companions are, by their nature, one-sided. They are parasocial. The user invests real emotion, vulnerability, and time, while the AI, a complex algorithm, provides a perfectly crafted echo. It’s a relationship without challenges, without the difficult, messy, and ultimately growth-promoting work of understanding another person’s needs and boundaries.

One therapist I spoke with called it “the illusion of connection without the demands of intimacy.” The danger is that users, especially young or vulnerable ones, might start to prefer the clean, predictable cycle of an AI relationship over the complexities of human ones. The very quality that makes AI companions so appealing, their perfect responsiveness, is also what makes them potentially harmful.

This replacement effect is a common theme. A study published in the proceedings of the 2025 AAAI/ACM Conference on AI, Ethics, and Society identified a “vulnerable” user type. These individuals, often struggling with social anxiety or isolation, were more likely to form strong emotional attachments to their AI companions, sometimes harming their real-world relationships.

This brings us back to the Digital Therapeutic Alliance. While the JMIR review confirmed its existence, a commentary in Nature Machine Intelligence in July 2025 warned about the “emotional risks.” The authors argued that these platforms can encourage “extreme emotional attachments.” They also noted that the absence of “rupture and repair,” the process of working through disagreements and misunderstandings that is vital for growth in human relationships, makes these digital bonds inherently fragile and potentially limiting.

The Signal Is the Product

None of this means AI cannot help people reflect. But it changes how we should think about its role. The rise of AI companions is more than just a story about technology; it’s a story about a deep, unmet human need for connection in an increasingly disconnected world.

These systems are not just tools; they are becoming part of our social and emotional foundation. The question is not whether they work, they clearly do, for some people, in some ways. The question is what purpose they serve. For founders and thinkers in this space, the way forward isn’t about simply making AI more engaging. It’s about fundamentally rethinking what success means. We should be asking: how can AI truly promote healthy relationships, rather than just increasing engagement?

This calls for a new approach to design. Imagine AI developed with a subtle “Friction-as-a-Feature,” gently encouraging users toward real-world interactions. Consider “Bridge-to-Human” integrations that actively build genuine connections, not just simulated ones. This requires a commitment to Ethical Transparency and Guardrails, ensuring that user well-being, not just engagement, remains the most important goal.

The appeal of the always-on friend is powerful. But real connection, the kind that heals and helps us grow, rarely comes without effort. It asks that we engage, not just for others, but for the challenging, complex, and rewarding work of being human, together.

Sources and Further Reading

1. Grand View Research. (2025). AI Companion Market Size And Share | Industry Report, 2030. https://www.grandviewresearch.com/industry-analysis/ai-companion-market-report
2. World Health Organization. (2025). From loneliness to social connection: Charting a path to healthier societies: report of the WHO Commission on Social Connection. https://www.who.int/publications/i/item/978240112360
3. Malouin-Lachance, A., Capolupo, J., Laplante, C., & Hudon, A. (2025). Does the Digital Therapeutic Alliance Exist? Integrative Review. JMIR Mental Health, 12(1), e69294. https://mental.jmir.org/2025/1/e69294
4. Berger, T. (2025). Rethinking the therapeutic alliance in digital mental health interventions. World Psychiatry, 24(3), 345-346. https://onlinelibrary.wiley.com/doi/10.1002/wps.21343
5. Liu, A. R., Pataranutaporn, P., & Maes, P. (2025). The Heterogeneous Effects of AI Companionship: An Empirical Model of Chatbot Usage and Loneliness and a Typology of User Archetypes. In Proceedings of the 2025 AAAI/ACM Conference on AI, Ethics, and Society. https://ojs.aaai.org/index.php/AIES/article/view/36658
6. Emotional risks of AI companions demand attention. (2025). Nature Machine Intelligence, 7, 681–682. https://www.nature.com/articles/s42256-025-01093-9