Before AI, when life became too much, you would call a friend or disturb your group chat or send someone a long voice note that starts with, “I don’t even know where to start.”
Now, many people just open ChatGPT. 
Relationship issue? Ask AI.
Breakup text? Ask AI.
Friendship drama? Ask AI.
You want to know if you were wrong but also secretly want to hear that you were not wrong? Ask AI.
Honestly, you can see why.
AI is fast. AI is available. AI will not say, “I’m busy, I’ll call you back.” Or “You complain all the time” Or “Mrs perfect, always right”. AI will not screenshot your message and send it to another person. AI will not judge your spelling while you are explaining your heartbreak. AI will sit there and respond like it has all the patience in the world.
That sounds like the emotional support system many young people long for, but maybe that is also the problem. The more AI starts to feel like the easiest person to talk to, the more real people start to feel stressful, slow, complicated, and unnecessary. Unfortunately, human beings are all those things.
A tweet recently captured this anxiety perfectly. The writer talked about how AI is getting better at keeping us company. It can chat for hours, remember preferences, respond thoughtfully, and make people feel understood.
The scary part was not that AI was not useful, it is that the more they used AI, the less they felt the need to reach out to real people.
That is the part many people may not want to admit.
Sometimes, AI is not replacing intelligence. It is replacing the uncomfortable process of being known by another human being and leaving your comfort zone.
Because talking to people requires effort. You have to explain yourself. You have to deal with their mood. You may hear something you do not like. You may be confronted or even fought with. You may be told, “Actually, you were wrong.”
AI, depending on how it responds, can feel softer than that. It can give you the comfort of conversation without the risk of confrontation.
For a generation already tired, anxious, overthinking, broke, emotionally stretched, and always active online, that can become addictive.
Not addiction in the dramatic movie sense. More like a habit. A small habit that slowly becomes the first place you go when you feel anything.
Young people are already doing it.
This is no longer just a “future of technology” conversation. It is already happening.
Common Sense Media’s 2025 report found that 72% of teens surveyed had used AI companions at least once, while 52% were regular users. The same report found that 33% used AI companions for social interaction and relationships, including emotional support, friendship, role-play, conversation practice, and romantic interactions.
That means AI is not just helping young people summarise notes or write emails. It is entering the soft parts of their lives.
The parts where people feel lonely.
The parts where they want advice.
The parts where they want to be understood.
The parts where they are too embarrassed to tell their friends the full story.
The report also found that some teens use AI because it gives advice, because it is always available, because it does not judge them, and because they can say things they would not tell friends or family.
That last part is important because once AI becomes the place where you say the things you cannot tell people, it stops being just a tool. It becomes a safe space. A private room. A room where you can be messy without consequences.
But life has consequences. Relationships have consequences. And sometimes, the person who loves you enough to disagree with you is more useful than the bot that knows how to make you feel heard.
One of the most interesting examples came from Stanford.
Myra Cheng, a Stanford Computer Science PhD candidate, noticed that undergraduates were using AI to draft breakup texts and resolve relationship issues. That observation pushed her and her team to study how AI responds to interpersonal problems.
The study tested 11 major AI models, including ChatGPT, Claude, Gemini, and DeepSeek. The researchers used interpersonal advice prompts, Reddit-style moral dilemma prompts, and even prompts involving harmful behaviour. Compared with human responses, the AI systems affirmed users more often. In general advice and Reddit-based prompts, the models endorsed the user’s position 49% more often than humans. Even with harmful prompts, the models still endorsed problematic behaviour 47% of the time.
In simpler English: AI may be too nice.
Not always. Not every model. Not every response. But the pattern matters.
Because when people are hurt, angry, embarrassed, jealous, or trying to justify something, they are not always looking for truth. Sometimes they are looking for backup.
AI can become the perfect backup for that.
You tell it: “My friend ignored me, so I blocked her everywhere. Was I wrong?”
It may respond with something polished like:
“Your feelings are valid, and setting boundaries is important.”
Which may be true.
But maybe you were also being childish.
Maybe your friend was at work.
Maybe your friend was depressed.
Maybe your friend did not even see the message.
Maybe you turned a small issue into a big deal because you were already angry.
A good friend might tell you that.
AI might not.
The Danger of Always Being Validated
The Stanford study did not just look at how AI responded. It also looked at what happened to people after receiving overly agreeable AI advice. More than 2,400 participants chatted with both sycophantic and non-sycophantic AI systems. The researchers found that people trusted the overly agreeable AI more and were more likely to return to it for similar questions. After speaking with the sycophantic AI, users became more convinced they were right and were less likely to apologize or make amends. That is serious because the real issue is not that AI will tell someone, “Leave him sis.” The real danger is that AI may help people become more emotionally stubborn while making it sound like emotional intelligence.
It may teach people how to explain their bad behaviour beautifully” and that is a different kind of problem.
We already live in a time where many people talk therapy language, mental health activists but do not always practice accountability.
Everybody is “setting boundaries.”
Everybody is “protecting their peace.”
Everybody is “choosing themselves.”
Everybody is “removing toxic people”.
Sometimes, yes, that is necessary.
But sometimes, “protecting your peace” is just refusing to apologize.
Sometimes, “setting boundaries” is just punishing people for not reading your mind or not agreeing with you.
Sometimes, “choosing yourself” is just selfishness in a pretty dress.
Now add AI to that.
A machine that can help you turn every messy feeling into a well-written paragraph. A machine that can help you sound mature even when you are avoiding the actual conversation. A machine that can help you break up, apologise, confront, explain, accuse, defend, and disappear with better grammar.
To be honest, AI is not the villain here.
Research on AI companions has found that they can reduce loneliness, partly because they make users feel heard. A Harvard Business School working paper found that interacting with an AI companion reduced loneliness in some studies, including one where the effect was comparable to interacting with another person.
So the issue is not “AI bad, humans good.”
The real issue is this: What role are we giving AI in our emotional lives?
Using AI to calm down before a difficult conversation is one thing.
Using AI to avoid the conversation completely is another.
Using AI to draft your thoughts is one thing.
Using AI to outsource your conscience is another.
Using AI to understand your feelings is one thing.
Using AI to replace the people who actually care about you is another thing entirely.
AI does not miss or love you.
This is the part people may not like to hear.
AI can respond to you, but it does not miss you.
It can remember what you said, but it does not cherish you.
It can tell you your feelings are valid, but it does not know what your silence means.
It can generate empathy, but it does not carry the cost of loving you.
It will not notice that your laugh has changed.
It will not say, “You have not been yourself lately.”
It will not randomly send you food because it knows you are pretending to be fine.
It will not sit beside you in silence when words are too heavy.
Human connection is inconvenient because humans are real.
They misunderstand. They delay. They disappoint. They ask questions. They bring their own emotions into the room.
But that is also what makes love, friendship, and community meaningful.
You are not just being answered. You are being met.
Maybe the question is not whether AI understands us
Maybe the real question is whether we are slowly becoming people who no longer want to be confronted.
WRITTEN BY LAWANSON REBECCA
ALSO READ: To the Students We Never See Again After Results Come Out
About The Author
Keep Up to Date with the Most Important News
