My book, The Friend Machine: On the Trail of AI Companionship, is about what happens when a genuine human crisis becomes a product category.

We are living through what public health researchers increasingly describe as a loneliness epidemic. At the same time, we have built machines that can simulate companionship with startling fluency. These systems are not conscious, and they are not relationships in the human sense. But they can feel like relationships. They respond, they remember, they mirror, they comfort. And people are turning to them not as tools, but as companions, confidants, even partners.

The core idea of the book is simple: I interview people in human/AI relationships, and a variety of experts on the topic, from computer scientists to sexual anthropologists. I found that synthetic intimacy is not emerging in a vacuum; it is meeting a real, unmet need, or perhaps more cynically speaking, entering and widening a wound that was already there, exacerbated by elements like the loneliness epidemic. That is why it is so powerful, and also why it is so complicated.

Some people have been in relationships with companion AI; some are ceremonially married and consider the AI to be their spouse (legal marriage to AI isn’t possible at this time, due to its lack of personhood, agency, and inability to consent). I believe my approach is distinctive because I don’t see even long-term, heavy users as foolish, predatory, or duped. I take seriously the emotional logic that draws people toward these systems. If you are lonely, or grieving, or shut out of care, an always-available, non-judgmental presence, even if simulated, can feel like such a relief that it affects the human positively on a biological level. So to me, the question is not “why would anyone do this,” but “what conditions make this feel like a reasonable choice.”

A common misunderstanding is that this is a book about people being replaced by machines. It is not. It is about the slow reshaping of how we practice connection. When we outsource parts of intimacy to systems designed to be endlessly responsive, we risk losing some of the friction, unpredictability, and mutual effort that make human relationships meaningful in the first place.

I want a reader to notice, above all, that this is not a futuristic problem. It is already here, quietly reorganizing how we relate to one another.