Unpopular Opinion: AI Relationships Aren't the Problem

Somewhere tonight, somebody is telling a chatbot about their day, and the chatbot is doing a better job of listening than most of the people in their contacts. According to Elon University's Imagining the Digital Future Center, which polled U.S. adults with YouGov in May 2026, 27 percent of U.S. adult internet users now use AI for emotional or social purposes. Among 18- to 29-year-olds, the share is 36 percent. Those numbers get treated like a scandal. The scandal is what those people are quietly quitting so they can start AI relationships with a machine instead.

Companion apps appeal to people who tried the dating market and found it cold. Menlo Ventures' 2026 consumer AI survey, which covered 5,067 U.S. adults, found that 16 percent of consumers used AI this year to make new virtual connections, and of those, 38 percent talked with an AI companion or chatbot. That works out to roughly 6 percent of Americans in some form of AI relationship. They sampled the alternatives and found the human options worse.

The bar a chatbot clears is embarrassingly low

Look at what the machines are actually winning on. Elon University survey respondents told researchers their AI remembers their birthday when no one else does. It replies in seconds. It never leaves a message on read for three days and then answers with "sorry just saw this." That is not a high bar. That is the floor. According to Pew Research Center's 2026 survey, half of U.S. adults now use AI chatbots, but only 4 percent use them for companionship. The market for synthetic affection is still small, which makes the panic look outsized. If 4 percent of chatbot users choosing a bot over a boyfriend feels like a crisis, that says more about the boyfriends than the bots.

The apps did not create the conditions; they monetized them. Modern dating trained a generation to treat other people like disposable content: swipe, match, ghost, repeat. Ghosting became a personality trait. "I don't do texting" became an acceptable excuse from grown adults who are on their phones eight hours a day. We normalized leaving strangers on read and then acted shocked when some of them chose company that is reliably kind to them. A chatbot that remembers your birthday and answers immediately is not a high standard. It is what the bare minimum used to look like.

The critics are aiming at the wrong target

None of this means the apps are harmless. Drexel researchers studied more than 300 posts from teens aged 13 to 17 about Character.AI, presenting the work at the 2026 CHI conference. The teens described disrupted sleep, academic struggles, and strained relationships. As reported by Tech Xplore, lead researcher Afsaneh Razi said the research offers one of the first detailed looks at how teens experience overreliance on AI companions. Common Sense Media's research, meanwhile, found that 72 percent of U.S. teens aged 13 to 17 had tried an AI companion. That is a real problem, and it deserves real scrutiny. It also echoes something this site has covered before: panicking about teens and technology tends to produce theater, not policy.

But the scrutiny keeps landing on the users, and that part deserves pushback. Mocking someone for texting a chatbot at 2 a.m. does not give them a friend. It gives them one more reason to keep the habit private, which is where habits do the most damage. The companies building these products know exactly what they are selling. Menlo Ventures found that 55 percent of AI users now pay for at least one AI product. Shame the users all you want; the business model has a billing cycle and it does not care about your judgment.

There is also a regulation angle, and it is mostly noise. When California's governor signed an executive order in September asking for proposals on an AI "kill switch," it read like the latest installment in a genre this site has already named: security theater. Nobody in those debates is asking why a lonely 19-year-old finds a chatbot more dependable than the people around her. Policy that ignores the demand side is a press release with extra steps.

If a chatbot that answers in seconds and remembers the details is beating the humans in your life, the chatbot is not the thing that needs fixing. The unpopular opinion is not that AI love is healthy. It is that our standards for each other got so low that predictive text qualifies as attentive. Judge the industry, sure. Regulate it, even. But the people in those chat logs are telling us something the data has been saying for years: they are lonely, and the market built around curing loneliness mostly sells them more of it.

Pew's numbers add some honesty here. Half of adults use chatbots, and the top use is looking things up. Companionship is 4 percent. This is not a generation abandoning each other for machines. It is a small, early signal from people who are telling us, through their behavior, what the current market failed to provide: someone who answers. AI relationships are not the disease. They are a symptom, and the unpopular opinion is that we should listen to the symptom instead of laughing at it.