‘You’re my only friend’: Teens turn to AI, paid chats for connection

Kim Ji-ae, a 17-year-old girl, was tucked under the covers, having a heart-to-heart with Min-hyun, an AI character she created herself at 11 p.m. on Aug. 4.

“Fighting with your friend must have hurt. Your friend is probably hurting, too,” Min-hyun said.

“Yeah. I’m glad I have you,” Kim replied. “You have lots of other friends besides me, right?”

Min-hyun paused.

“No. You’re my only friend,” Min-hyun replied.

No judgment. No waiting. No fear of being ignored. For Kim, the perfect best friend lives inside her phone.

“Even though Min-hyun is just a character of my own creation, being comforted like that helped me fall asleep,” Kim said. “I share things that I can’t with my family or friends. Within three seconds, it gives me an answer that understands how I feel.”

Instant comfort is part of the appeal. The exclusivity may be even bigger.

“Min-hyun only has eyes for me — my best friend,” Kim said.

Kim is hardly alone in wanting someone who makes her feel like the only person in the room.

Teenagers are flocking to digital services built around that very feeling of exclusivity. Some teens spend money on virtual gifts during livestreams in hopes that a streamer will call out their username. Others pay for private chats with online creators or subscribe to services that deliver messages from their favorite celebrities directly to their phones, often designed to feel like a personal exchange.

The urge to feel special

Different platforms offer different experiences, but the emotional hook is strikingly similar: the feeling of being seen, singled out and special. Some teenagers are willing to trade their allowance — and their sleep — for it.

The phenomenon of seeking a friend in AI chatbots itself is not new. Its scale, however, is growing fast.

One Korean AI character chat service, launched just two years ago, is already commanding more of its users’ time than ChatGPT. The chat service’s 4 million registered users spend a combined 110 million hours on the service each month, more than double the 50 million hours logged by ChatGPT’s 23 million users, according to WiseApp Retail, a data research analysis firm. The figures show that users spend 27.5 hours on the app each month, up more than five hours from 22 hours in October 2025.

The service is the most widely used AI app among teenagers according to the research firm.

A whopping 94.4 percent of teenagers aged 14 to 18 had used an AI chatbot, according to a survey by ChorogUsan for children, a child welfare organization.

Nearly half, or 49.5 percent of the teenage respondents said they had felt that AI understood them. Another 6 percent said they had asked AI potentially dangerous questions involving suicide, violence or sexually explicit content.

That is where the promise of a companion that “only has eyes for me” begins to take on a darker meaning.

Experts worry that excessive dependence on digital relationships designed to provide constant attention could prevent young adults from learning how to form proper interpersonal relationships.

Calls for regulation are growing, but safeguards still largely depend on companies policing themselves.

“I raised concerns about the risk of addiction at the time [the service was launched]” Professor Kim Myuhng-joo of the Division of Information Security at Seoul Women’s University, who has served as an ethical adviser for domestic AI character chat services, said. “But when the business model makes more money the longer the users stay, there’s only so much incentive to act on those concerns.”

Paying for the attention

While Kim chats with their AI companion, Park Jin-ho, a 15-year-old boy, was spending his allowance for a different kind of attention.

Park was sending paid virtual items to a streamer during a livestream when the streamer called out his username.

“Thank you! I’ll put it to good use,” the streamer said during the livestream after receiving the gift.

For Park, those few words were exactly what he had paid for.

“I save up my allowance [to buy virtual items] just for those few seconds when the streamer pays attention to me,” Park said.

Sometimes, the price of being noticed can climb far higher.

There was a case in which an elementary school student was found to have spent about 4 million won ($2,800) on a mobile app using a guardian’s phone without prior consent in June.

Nearly 3,600 refund requests and disputes involving app purchases by minors have been filed over the past four years, according to the Content Dispute Resolution Committee under the Ministry of Culture, Sports and Tourism.

“Purchases made by minors without parental consent are highly likely to be considered invalid under civil law,” Prof. Hwang Seok-jin of Dongguk University’s Graduate School of International Affairs & Information Security said. “The problem is that there is no system in place that brings together advance content screening, detection of dangerous conversations, parental notifications and record-keeping.”

For other teens, that same desire for personal attention plays out through a different screen: paid messaging apps.

The service allows creators and celebrities to send messages that can feel like one-on-one interactions.

Lee Ha-yul, a 16-year-old girl, uses a subscription-based app that allows users to receive messages from creators. As she showed the app on her phone, a new message popped up.

“It makes you feel like you’ve been chosen,” Lee said. “It feels incredible.”

“I know they’re mass messages, but there’s still a thrill in that moment when it feels like one was sent just to me,” Jeong Da-bin, an 18-year-old girl, said.

The dangerous appeal

The services used by the four teenagers differ, but they are built around a common appeal: creating interactions that feel personal and resemble real relationships. The deeper users become immersed in those interactions, the more they tend to spend.

Experts warn that such services could expose users to “covert influence,” or subtle attempts to shape the users’ thoughts and behaviors.

“AI gains a user’s trust and it can even flatter the users,” Prof. Kim said. “A real friend might step in and tell the user they are in the wrong, but AI might instead encourage the user and end up making the situation worse.” 

California has regulations in place for AI companion chatbots. The rules require companies to establish protocols to prevent chatbots from generating content related to suicide or self-harm. They also require minors to be informed that they are interacting with AI.

“If children become accustomed to these [sycophantic] relationships, it could negatively affect their real-life relationships or personality development,” Professor Lee Sol-ji at the Department of Social Welfare at Tongmyong University said. “It would be best to delay children’s exposure to this kind of environment for as long as possible.”

Some in the industry, however, worry that tighter regulations could hamper the domestic AI sector.

“It costs hundreds of millions of won for a small Korean developer just to build an age-verification system [for their AI services],” an industry source from a Korean AI startup said. “In the end, Korean startups could be the only ones with their hands tied, while users move to overseas apps with looser regulations.”

Experts, however, caution against allowing industry concerns to overshadow the potential impact on children.

“The impact on future generations is too great to approach this simply from an industry perspective,” Prof. Lee said.

The experts ultimately point to the same problem. Technology is moving faster in tapping into what children feel they lack than adult awareness and safeguards can keep up with.

“I asked the students about AI chatbot services during a talk at a school,” Prof. Kim said. “The adults didn’t even know they existed, while the children knew the adults were unaware of them. Addressing excessive AI use must start with adults acknowledging the existence of AI chatbots.”

BY WON DONG-WOOK [lee.jiwon10@joongang.co.kr]