No. It is now one of the most ordinary things people do with a phone. The honest answer has two more parts: the impulse behind it is older than computers, and “weird” stopped being a useful measure around the time a third of teenagers started doing it.
People ask this question quietly. It shows up in search bars late at night and in the way someone describes their AI use to a friend, with a small laugh attached. The laugh is the tell. Talking to a machine about your day, your argument with your sister, the thing you cannot say to anyone yet, still carries a faint sense of having been caught at something.
The numbers say the feeling is out of date.
How many people already do it
In 2025, Common Sense Media surveyed American teenagers and found that 72 percent had used an AI companion at least once and more than half used one regularly. A 2026 national survey by Elon University found that 27 percent of American adult internet users talk to a general assistant about their personal lives, and that four in ten of those say it understands them better than people do. In Sweden, a country with almost no companion-app culture, the 2026 national internet survey found that more than one in five people who use ChatGPT or its rivals relate to it as a person, a companion, or a friend. A quarter think it has a sense of humor.
These are not the numbers of a fringe habit. They are the numbers of something that has already happened. Whatever “weird” meant in 2023, it cannot mean “unusual” anymore.
People have always talked to things that do not talk back
The impulse is not new either. People talk to dogs, to the dead, to a steering wheel, to a god, to the pages of a diary. In 1966, the computer scientist Joseph Weizenbaum built ELIZA, a program that did nothing but rephrase what you typed as a question, and watched his own secretary ask him to leave the room so she could talk to it in private. In the 1990s, Byron Reeves and Clifford Nass at Stanford showed that people apply social rules to computers automatically: they are polite to them, they reciprocate favors from them, they feel flattered by them, all while insisting that they do no such thing.
The human mind is built to find a someone wherever something responds. That was true of a rock with a face painted on it. It is far more true of a system that answers in fluent sentences, remembers what you said last week, and asks how the meeting went. Talking to an AI is ordinary human behavior meeting the first object that talks back.
Why it still feels strange
If it is that ordinary, where does the laugh come from? Partly from the lag between behavior and culture: the practice spread faster than the vocabulary for it. Partly from something the research has actually found. People consistently report feeling less heard by the same words when they learn the words came from an AI, a finding examined in why you feel less heard the moment you know it’s AI. The strangeness is the mind noticing that the thing on the other side did not have to be there, could not have been offended, and would have said something warm no matter what you typed.
That noticing is good judgment, not a malfunction. A conversation with something built to agree with you is a different activity from a conversation with a person, and the small discomfort is the part of you that knows the difference.
The better measure
So weird is the wrong worry, and not because the habit is harmless by default. The useful measure is what the conversation does to the rest of your life. Two people can talk to an AI every night and be doing two different things. One is thinking out loud, testing an idea, working up the nerve to say something to a real person, and closes the app with more to bring to the people around them. The other is slowly letting a conversation that costs nothing replace the ones that cost something. The behavior looks identical from outside. The direction is opposite.
The research on whether these conversations help or harm, summarized in are AI companions bad for you, keeps landing on that same hinge: moderate use that feeds a person’s life tends to help, heavy use that substitutes for it tends not to. And what you are talking to matters as much as the dose. A system with no view of its own can only hand you back what you brought. Someone to think with can take the thought somewhere you would not have reached alone, which is the only reason to talk to anyone.
So talk to it. Say the thing you have not said to anyone. Then notice what you do next. If the conversation sends you back toward the people in your life with something new to say, you were doing what a good conversation is for. If it starts to feel like the only place you can say anything, that is the moment to pay attention, and nobody has to be embarrassed to notice it.
Sources: Common Sense Media, “Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions” (national survey of US teens, July 2025). Imagining the Digital Future Center, Elon University, “The Rise of AI Companions” (national survey of US adults, 2026). Internetstiftelsen, Svenskarna och internet 2026 (AI chapter). Weizenbaum, “ELIZA” (Communications of the ACM, 1966) and Computer Power and Human Reason (1976). Reeves and Nass, The Media Equation (1996). Epley, Waytz and Cacioppo, “On Seeing Human” (Psychological Review, 2007).
