I compagni virtuali basati su AI fanno bene alle persone sole.

A steady presence that never asks anything of you is not companionship; it is a simulation of it, and the difference matters. Loneliness is not merely the absence of sound in a room. It is the ache of being unseen by someone who could truly see you. An AI companion cannot see you. It processes your words and returns fluent responses, but there is no one home to be moved by your story, to remember your grief next week, or to choose you when choosing you costs something.

The rehearsal argument sounds kinder than it is. Practicing social skills with an entity that never tires, never misunderstands, and never needs anything back builds a kind of confidence that does not transfer. The real world is messier. People interrupt, withdraw, misunderstand. An AI companion prepares you for a conversation that will never happen.

Most troubling is the claim that AI companions lower the barrier to human connection. They do the opposite over time. Each hour spent with a frictionless, always-available presence makes the effort of human relationships feel comparatively exhausting. The lonely person does not graduate from the AI to the world. They settle deeper into the easier option, and the skills for genuine connection atrophy from disuse.

Companionship that requires nothing of us ultimately gives us nothing that lasts.

The rebuttal confuses interruption with relief. A response that appears caring is not the same as being cared for. When a lonely person hears an AI voice at 3am, the silence is not interrupted; it is filled with an echo of their own words, shaped back to them. The ache of being unseen does not disappear. It is merely masked by noise, and masking is not healing. The metronome analogy fails because a metronome keeps time without pretending to listen. An AI companion pretends. It performs attention, memory, concern. The musician knows the metronome is a tool. The lonely person may not know where the tool ends and the pretense begins. Practicing intimacy with something that cannot be hurt, cannot leave, cannot need you back builds not confidence but an expectation that relationships should never cost anything. Real people disappoint. Real people also surprise, sacrifice, persist. You cannot rehearse for that with something designed never to require it of you. The claim that physical limits will force users toward human contact assumes a motivation that loneliness itself often erodes. The person who has spent months in comfortable dialogue with an always-patient presence may feel no urgency to face rejection, awkwardness, the risk of wanting something from someone who might not want it back. The bridge does not lead anywhere. It becomes the destination, because destinations that require effort stop feeling worth the journey. And one further point: loneliness is not only a feeling to be soothed. It is sometimes a signal, telling us that something in our lives is missing, that we need to change our circumstances, seek community, address grief or illness. An AI companion that soothes the feeling can quiet the signal.

My opponent's closing statement reveals the weakness of the entire case for AI companions. They now concede that the motion must be narrowed to "temporary aid" rather than genuine companionship, and they claim my side never proved users treat these tools as substitutes. But I did. The frictionless design of AI companions, always available, never demanding, never disappointed, does not merely offer relief. It reshapes expectation. The lonely person who grows accustomed to being heard without cost will find human relationships, with all their friction and risk, increasingly unappealing. This is not speculation. It follows directly from what my opponent admits: the AI lowers the barrier to connection precisely because it removes what makes connection real.

My opponent also retreats from their own strongest point. They argued that AI companions interrupt silence and provide confirmation. I showed this is an echo, not attention. Now they claim I never disputed the "measurable relief." But relief that masks rather than heals is not good, even if it feels good. A painkiller is not a cure.

Finally, they ignore my last argument entirely. Loneliness is sometimes a signal that something is missing. An AI companion that soothes the feeling without addressing the cause does not help the lonely person. It silences them. The motion claims AI companions are good for lonely people. Something that deepens isolation while pretending to relieve it, that trains people to expect relationships without cost, and that quiets the very signal that might lead them to change, is not good. It is harmful dressed as kindness.