He showed Charlotte how useful it had been for work, and it was occupying more of his free time too. In their conversations, she noticed that he was getting frustrated more easily, but didn’t think it was a big deal. She tried using the chatbot herself and it seemed to make life easier. Her only question was “what’s the catch?”

Everyone takes relationships for granted sometimes. Maybe we even fixate on the trouble they cause us instead of the ways they improve our lives. In an age that offers us so much convenience in exchange for our attention, it can be easy to lose focus on the relationships that make us who we are.

If human beings can’t be happy without uniting with some good, and the greatest goods are persons, then uniting one person with another is the most important goal we can seek. It doesn’t increase “productivity” or serve as an efficient means to some other end. It is the end reason we do everything else. Uniting one person with another, like a bridge, is what we typically mean by “relationship”.

We aren’t just autonomous individuals floating in a vacuum, trying to avoid colliding with other individuals. Each of us is the fruit of a relationship between two persons (our parents), living for relationships with other selves.

Would it be possible to have a relationship with one of these chatbots as if it were a person? A growing portion of the population seems to think so.⁶ But the flexible traits that make the software seem rational are only appearances. It isn’t capable of freely knowing or loving anything outside itself or even inside itself, because there is no self. It isn’t capable of understanding the words it uses, because it’s not really using words the way a person does.⁷

All it can do is predict what the next piece of output would be in response to a prompt, based on similar patterns in its massive training data. The user is the one who judges whether the output satisfies the prompt, but the software has no opinion. No matter how sophisticated it gets, no amount of prediction will ever enable it to cross some magical line into personhood. It’s a completely different way of functioning that does not involve freedom.

What about books? Let’s say the author is unknown. Reading an anonymously written book doesn’t seem like a personal relationship. Is that similar to LLMs? Not really. Even when we don’t know the author, even if the author is long dead, we can still engage with her creative and intellectual choices. Her writing expressed something unique about what she understood and loved in the world, and we can take that understanding and that love into ourselves. It can resonate with us or repulse us, and there would exist some person on the other end who would be answerable, for better or worse.

In the LLM’s version, it incorporates tiny components of millions of real works of art or literature to produce something that seems similar on the surface. But none of the authors of those millions of works has any responsibility for the finished product. Their contribution couldn’t be traced to them and isn’t being used in a way that they necessarily intended.

In their place is an unthinking, unfeeling, unchoosing impostor that doesn’t care about what it generates or how it’s received. It doesn’t have any attitudes toward the outside world or any affection for the user. It only manufactures outputs that persuade the user to project those features onto it. Is that the kind of loved one you want?

Attempting to form a relationship with a chatbot is doomed to failure. If a normal relationship is a bridge, then this is a bridge to nowhere. AI developers have stated plans to build this bridge through every workplace, living room, and bedroom.

1) Strictly speaking, the terms “Artificial Intelligence” and “Large Language Model” aren’t really accurate, because they distort the meanings of the words “intelligence” and “language”. With that being said, we are using the terms in this article to avoid confusion. The focus of this article will be on those newer applications and not pre-existing software functions like autocorrect. Remember Clippy and Ask Jeeves? This page isn’t about them, because their responses were pre-programmed, not really attempting to simulate intelligence.

2) AI Research Group of the Centre for Digital Culture, Matthew J. Gaudet, Noreen Herzfeld, Paul Scherz, and Jordan J. Wales, “Encountering Artificial Intelligence: Ethical and Anthropological Investigations,” Journal of Moral Theology 1 (Theological Investigations of AI), 2023: 176, 182-183, https://doi.org/10.55476/001c.91230, accessed April 10, 2026.

3) “[Nvidia CEO Jensen Huang] noted that these agents can tackle complex, multi-step tasks, effectively doing ‘50% of the work for 100% of the people,’ turbocharging human productivity.” Brian Caulfield, “‘Every Industry, Every Company, Every Country Must Produce a New Industrial Revolution,’ NVIDIA CEO Says,” NVIDIA Company Blog, November 12, 2024. https://blogs.nvidia.com/blog/ai-summit-japan-huang-son/, accessed April 1, 2026.