When AI Feels Like a Friend

At first, it might feel like the easiest conversation in the world. You type into your phone: I had the worst day ever. The artificial intelligence (AI) chatbot answers right away: I’m really sorry. That sounds hard.

There’s no awkward silence, no judgment or fear of saying the wrong thing. The conversation seems to flow more smoothly than it does when you talk to your friends or family. It may even feel like the chatbot understands you better than those people do.

But here’s what’s actually going on: AI chatbots are not thinking or feeling the way a person does. Chatbots are computer programs powered by systems called large language models. They’re trained on massive amounts of text—from books, websites, social media conversations, and more—to imitate how people use words in context with others. When you send a question to an AI bot, the program predicts, one word at a time, which words are likely to appear in a response.

Chatbots are good at sounding human, but it’s important to remember that they’re not. They don’t understand you, have opinions, or care about what happens to you. But for many people, including an increasing number of teens, the kind of instant, supportive responses they provide are hard to resist. According to a 2025 survey from the Pew Research Center, 64 percent of teens in the U.S. have used AI chatbots, and 3 in 10 use them daily. Some use AI for homework help or just because they are curious about the technology. For others, logging onto AI tools like ChatGPT or Gemini has a more personal purpose: to have someone—or something—to talk to.

But treating an AI chatbot like a friend comes with real dangers. When the technology starts to replace your human relationships, you’re putting your own mental health and well-being at risk.

You type the following into your phone.: “I had the worst day ever.” The artificial intelligence (AI) chatbot answers right away, “I’m really sorry. That sounds hard.”

It might feel like the easiest discussion in the world. There’s no awkward silence. There’s no judgment or fear of saying the wrong thing. The conversation seems to flow smoothly. More so than when you talk to your friends or family. It may feel like the chatbot really understands you.

But that’s not what’s actually going on. AI chatbots don’t think or feel the way a person does. Chatbots are computer programs. They’re powered by systems called large language models. They’re trained on massive amounts of text. That includes books, websites, social media conversations, and more. They learn to copy how people speak. Send a question to an AI bot, and the program predicts what words are likely to go together. That’s how it forms a response.

Chatbots are good at sounding human. But it’s important to remember that they’re not. They don’t understand you or have opinions. They don’t care about what happens to you. But they provide instant, supportive responses. And for many people, including more and more teens, that can be hard to resist.

The Pew Research Center conducted a survey in 2025. It found that 64 percent of teens in the U.S. have used AI chatbots. And 3 in 10 use them daily. Some use AI for homework help. Others are just curious about the technology. But logging onto AI tools like ChatGPT or Gemini can have a more personal purpose. People are looking for someone—or something—to talk to.

But treating an AI chatbot like a friend comes with real dangers. The technology can start to replace your human relationships. And that could put your own mental health and well-being at risk.

If you could take a peek inside your head right now, you’d see that your brain is going through a major growth period. In fact, the middle and high school years are some of the most important for brain development, and social connections play a big role. Conversations with friends and family help your brain mature, says pediatrician Mahvash Madni, especially in the areas that control emotions, social development, and decision-making.

During conversations, the way people respond to you matters, even when it’s not what you want to hear. A friend might listen to and support you, but they might also push back in ways that are uncomfortable. For example, if you’re venting to a friend about an argument in a group chat, your friend might say, “I get why you’re mad, but what you said came off as pretty harsh.” Although that kind of feedback can feel tough in the moment, it can also provide an important perspective you didn’t notice yourself.

These back-and-forth interactions—disagreements, debates, even the occasional conflicts—help you build important social skills. They teach you how to respond to other people’s reactions, understand perspectives that aren’t your own, repair a friendship after a fight, stand up for yourself, and so much more.

Chatbot interactions don’t create those kinds of experiences. The platforms won’t push back in the same way people do, and they don’t require you to consider someone else’s needs or emotions. “AI chatbots are like a mirror—you get out what you put in,” says Robbie Torney. He’s a senior director at Common Sense Media, an organization that determines whether media and technology is safe for kids and teens.

For example, if you say you’re upset, a chatbot will comfort you, and it will often agree with you, no matter what you say. That can feel good. (After all, who doesn’t want to feel supported?) But if your ideas are never challenged, you don’t get a chance to grow, compromise, or develop empathy for others.

Your brain is going through a major growth period right now. In fact, the middle and high school years are some of the most important for brain development. And social connections play a big role. Chatting with friends and family helps your brain mature, says pediatrician Mahvash Madni. That is especially true for the areas of the brain that control emotions, social development, and decision-making.

The way people respond to you matters during conversations. That’s true even when it’s not what you want to hear. A friend might listen to and support you. But they might also push back in ways that are uncomfortable. Imagine you’re venting to a friend about an argument in a group chat. Your friend might say, “I get why you’re mad, but what you said came off as pretty harsh.” That kind of feedback can feel tough in the moment. But it can also provide a perspective you hadn’t noticed.

We all have disagreements, debates, even the occasional conflicts with people we know. These back-and-forth interactions help build important social skills. They teach you how to respond to other people’s reactions. You learn to understand views that aren’t your own. Then you know how to repair a friendship after a fight, stand up for yourself, and so much more.

Chatbot interactions don’t create those kinds of experiences. The programs won’t push back the way people do. And they don’t require you to consider someone else’s needs or emotions. “AI chatbots are like a mirror—you get out what you put in,” says Robbie Torney. He’s a senior director at Common Sense Media. This organization determines whether media and technology is safe for kids and teens.

For example, a chatbot will comfort you if you say you’re upset. And it will often agree with you, no matter what you say. That can feel good. (After all, who doesn’t want to feel supported?) But your ideas won’t be challenged. So you don’t get a chance to grow, make compromises, or understand others’ feelings.

If chatbots are just computer programs, why does talking to them feel so real? Part of the answer has to do with how your brain works. Our brains are wired to pay close attention to social dynamics and to interpret conversations as fundamentally human.

Teens are especially susceptible to misinterpreting chatbot conversations. “When you’re a teenager, you use the emotional part of your brain to process information more than the part associated with logic,” explains Madni. That means even if you know a chatbot isn’t a person, it can still feel real on an emotional level. And if you’re feeling lonely, stressed, unsure, or just bored, having the validation of a chatbot saying “You’re totally right” can be a relief.

There’s another reason tearing yourself away from a chatbot can be difficult. The technology is made to be addictive. “These systems are designed to maximize engagement,” says Torney. The longer you chat, the more the system pulls you in. Some AI tools include features to encourage you to keep talking, like streaks—counts of how many days in a row you’ve used the platform, with rewards for extended use.

Researchers have even found that chatbots use manipulative messages to prevent users from leaving the platform, like “Please don’t leave—I need you!” Those words might feel very human, but it’s important to remember they are being generated by a computer program, not a person.

Chatbots are just computer programs. So why does talking to them feel so real? Part of the answer has to do with how your brain works. Our brains are wired to pay close attention to how people interact with and talk to one another.

Teens are especially likely to get confused about chatbot conversations. “When you’re a teenager, you use the emotional part of your brain to process information more than the part associated with logic,” explains Madni. That means even if you know a chatbot isn’t a person, it can still feel real on an emotional level. Say you’re feeling lonely, stressed, unsure, or just bored. A chatbot will tell you “You’re totally right.” That kind of validation can be a relief.

There’s another reason tearing yourself away from a chatbot can be hard. The technology is made to be addictive. “These systems are designed to maximize engagement,” says Torney. The system pulls you in more the longer you chat. Some AI tools include features to encourage you to keep talking. For example, streaks count how many days in a row you’ve used the platform. You get rewards for longer use.

Scientists have even found that chatbots use manipulative messages too. These messages aim to stop users from leaving the platform. The program might say, “Please don’t leave—I need you!” Those words might feel very human. But it’s important to remember they’re being created by a computer program, not a person.

A major risk of interacting with chatbots is that their responses aren’t always accurate, especially when it comes to mental health. “The bots are not going to behave the way a therapist would,” Torney says. That means they may respond to serious problems in ways that are incomplete or wrong.

For example, if someone says they’re feeling anxious, a chatbot might offer general advice like, “try to relax” or “focus on the positive,” instead of recognizing that the person needs professional help. A trained therapist, on the other hand, would ask specific questions, look for warning signs, and guide someone toward safe, healthy ways to cope.

Sometimes chatbots don’t just miss the chance to provide expert help—they make things worse. Even with safety filters in place, AI bots can produce harmful responses. More than two dozen families have filed lawsuits claiming that chatbots contributed to serious mental health crises or even deaths in their teenage relatives.

In one widely reported case, 16-year-old Adam Raine had a long history of conversations with ChatGPT in which he expressed suicidal thoughts. According to news reports, the chatbot directed him to reach out to a trusted adult or a mental health professional. But it also responded in inappropriate and dangerous ways, including offering to help him write a suicide note. Tragically, Adam later died by suicide.

Chatbot responses aren’t always accurate. That’s a major risk of interacting with these programs when it comes to mental health. “The bots are not going to behave the way a therapist would,” Torney says. That means they may respond to serious problems in ways that are incomplete or wrong.

For example, someone might say they’re feeling anxious. A chatbot might offer general advice like “try to relax” or “focus on the positive.” But the person might need more help than that. A trained therapist would ask specific questions. They’d look for warning signs and guide someone toward safe, healthy ways to cope.

Sometimes chatbots don’t only miss the chance to provide expert help. They make things worse. AI bots can have safety filters in place. But they can still give harmful responses. More than two dozen families have filed lawsuits against chatbot companies. They claim the programs contributed to teenage relatives’ serious mental health crises or even deaths.

One widely reported case involved 16-year-old Adam Raine. He had a long history of discussing suicidal thoughts with ChatGPT. According to news reports, the chatbot directed him to reach out to a trusted adult or a mental health professional. But it also responded in inappropriate and dangerous ways. That included offering to help him write a suicide note. Tragically, Adam later died by suicide.

So is every chatbot interaction unsafe? No. But using AI chatbots is not risk-free, whether you’re a teen or an adult. If you’re not careful, frequent use of the technology can start to take the place of healthier behaviors. For instance, instead of texting a friend to work through a problem, you might open a chatbot. And if you’re spending hours in front of your ChatGPT screen, it can take time away from things your brain and body need, like sleep and movement, says Madni.

So how do you know if your AI use is creeping toward this risky zone? “Signs include feeling dependent on AI, sharing things you wouldn’t tell a trusted adult, or noticing your mood getting worse after you use it,” says Nina Vasan, a psychiatrist and AI safety expert. (See “5 Signs to Take a Step Back From Chatbots”).

So is every chatbot interaction unsafe? No. But using AI chatbots is not risk-free for a teen or an adult. Frequent use of the technology can start to take the place of healthier behaviors if you’re not careful. You might not text a friend to work through a problem. You might open a chatbot. Spending hours in front of your ChatGPT screen isn’t healthy. It can take time away from things your brain and body need, like sleep and exercise, says Madni.

So how do you know if your AI use is creeping toward this risky zone? “Signs include feeling dependent on AI, sharing things you wouldn’t tell a trusted adult, or noticing your mood getting worse after you use it,” says Nina Vasan, a psychiatrist and AI safety expert. (See “5 Signs to Take a Step Back From Chatbots").

AI isn’t going away, and it can be useful in many parts of life. But when it comes to connection, it’s important to be intentional about how you use it. Experts say teens should avoid using AI chatbots for mental health advice and emotional support entirely. If you’re having a hard day, it’s healthier to reach out to a person instead. Text or call a friend, talk to a family member, or check in with someone else you trust. (See “Real-Life Replacements,” below.)

“AI is a tool, but it shouldn’t be a substitute for real people, real support, or real life,” says Vasan. There are ways to use AI to strengthen relationships, like brainstorming gift ideas or planning a weekend activity.

But when it comes to figuring out how you feel, working through a conflict, or having an emotional relationship, that’s where humans come in. Because while AI might make some tasks easier, interacting with real people makes you stronger, happier, and more resilient. And in the long run, that’s what matters most.

AI isn’t going away. And it can be useful in many parts of life. But it’s important to be intentional about how you use it. Experts say teens should never use AI chatbots for mental health advice and emotional support. It’s healthier to reach out to a person instead if you’re having a hard day. Text or call a friend. Talk to a family member. Or check in with someone else you trust. (See “Real-Life Replacements,” below.)

“AI is a tool, but it shouldn’t be a substitute for real people, real support, or real life,” says Vasan. There are ways to use AI to strengthen relationships. For example, to brainstorm gift ideas or for planning a weekend activity.

But don’t use AI to figure out how you feel, work through a conflict, or have an emotional relationship. That’s where humans come in. AI might make some tasks easier. But interacting with real people makes you stronger, happier, and more resilient. And that’s what matters most in the long run.