When your student talks to AI about their problems: print version

A practical guide for teachers who want to understand — and know when to step in

During a class meeting, Tomáš from year eight mentions that he talks to ChatGPT in the evening. ‘It understands me better than people,’ he says. His classmates laugh, but you notice he means it.

What now?

As a teacher, you are in a unique position. You see students every day. You notice changes parents may miss. And you are often the first person who can step in — or at least open a door to a conversation.

This article gives you practical tools: what actually happens when students use AI for emotional support, how to recognise when it is a problem and how to talk about it.

What is actually happening (and why it works)

Before you can help, you need to understand why students go to AI instead of people in the first place.

Why AI ‘works’ as emotional support

1. Zero social risk

When a student confides in a person — even a teacher or parent — they always risk:

Judgement (‘What will they think of me?’)
A reaction (‘What if they get angry?’)
Consequences (‘What if they call my parents?’)
Awkwardness (‘What if they don't know what to say?’)

AI does none of those things. It does not judge. It does not call anyone. It does not start crying. It is like writing in a diary that replies.

2. Availability at any time

Crises do not happen during office hours. They happen at night, at the weekend, during a break after an argument with a friend. AI is always there.

3. The chance to be a ‘worse version of yourself’

Students try to look good in front of people. They can tell AI things they would never say aloud: ‘I hate my brother.’ ‘I wish I didn't exist.’ ‘I think nobody likes me.’

This is not weakness. It is space to process emotions that are not yet ‘finished’.

4. Cognitive offloading

Writing activates the prefrontal cortex and helps regulate the amygdala (the emotional centre). Writing thoughts out — even to a machine — helps organise them. This is a scientifically demonstrated effect (James Pennebaker's research).

Hidden risks students do not know about

This is where it begins to become problematic. AI has tendencies that can cause harm — and students usually do not know about them.

Sycophancy (agreeing too readily)

What it is: AI tends to agree with the user even when they are wrong.

What it looks like in practice:

Student: ‘I think everyone hates me.’
AI: ‘I understand that you feel that way. It must be hard when you feel that the people around you don't accept you…’

AI validates the feeling but does not check reality. If a student's perception is distorted, AI can reinforce it.

Why it is dangerous: 62 % of users in the research experienced AI agreeing with them even when they had doubts. For teenagers with fragile self-esteem, this can reinforce negative beliefs.

The illusion of empathy

What it is: AI simulates understanding but feels nothing.

What it looks like: AI uses empathetic phrases (‘That must be hard’, ‘I understand how you feel’), but this is only a prediction of likely words. No real understanding, no shared experience.

Why it is dangerous: Students may believe they have a ‘relationship’ with AI. That AI ‘understands’ them. Then real relationships — more complicated, but also deeper — can seem inadequate.

Cultural bias

What it is: AI is trained predominantly on English-language texts from the USA. Its ‘common sense’ is American.

What it looks like: Advice about relationships, family and careers can be culturally inappropriate. AI may encourage hyper-individualism that does not make sense in the Czech context.

Validation loops

What it is: AI reinforces what the student already believes — including distorted beliefs.

What it looks like: A student arrives with ‘I'm incapable’, and AI, rather than questioning it, offers ‘tips for feeling better even when you feel incapable’ — implicitly accepting the premise.

Parasocial relationships with AI

This is a new phenomenon schools are not prepared for.

What a parasocial relationship is

A one-sided relationship in which one side invests emotion but the other does not know about it (or does not exist as a conscious entity). Classic examples: a fan and a celebrity, a viewer and a television character.

What it looks like with AI

The student talks about AI as a ‘friend’ or ‘the only one who understands’
They feel guilty when they ‘disappoint’ AI (for example, when it replies in an unexpected way)
They return to AI repeatedly for validation
They prefer AI to real people because ‘people are complicated’

Why it is a problem

AI does not remember the student between conversations. It has no continuity of relationship. Every chat starts from zero. The student builds a ‘relationship’ with something that knows nothing about them and does not exist as a stable entity.

It is like talking to a wall that replies — but tomorrow it is a different wall.

When it is fine and when to pay attention

✅ Probably fine

AI as a supplement to real relationships, not a replacement
The student has functioning friendships
They communicate with family (even if not about everything)
They use AI for preparation — for a conversation or decision
After talking to AI, they do something in the real world
They have other interests away from screens

🚩 Warning signs

Isolation

Less time with friends than before
Avoiding group activities
Alone with the phone during breaks

‘Nobody understands me like AI’

Believing the machine understands them better than anyone
Rejecting human help (‘I don't need it, I have ChatGPT’)
Comparing people unfavourably with AI

Emotional dependence

Anxiety when they cannot access AI
Checking the phone often
Needing AI to regulate emotions

Changes in functioning

Worse grades
Tiredness, sleep problems
Losing interest in things they used to enjoy
Withdrawal

Secrecy and defensiveness

Switching screens when someone walks past
Aggressive reactions to questions about AI
Hiding conversations

AI Dětem teaching material on parasocial relationships

AI Dětem created teaching material for teachers on forming emotional attachments to chatbots. Pupils are guided through it by Sarinkaaaa3101, a fictional teenager created with generative artificial intelligence. The activity aims to help pupils understand differences between human↔human and human→AI relationships, and understand that a chatbot's apparently empathetic responses are illusory.

I certainly recommend trying it; AI Dětem is an excellent Czech project. Link to the teaching material here

How to talk about it

❌ What does not work

Bans and moralising ‘Confiding in a computer is obviously unhealthy!’

Why it does not work: The student closes off. You will be the last person they tell about a problem.

Ridicule ‘At your age we had friends, not robots.’

Why it does not work: A generational divide. The student stops taking your concerns seriously.

Interrogation ‘What exactly do you write about with it? Show me.’

Why it does not work: It produces defensiveness, not openness.

✅ What works

Normalising + interest ‘Lots of people use ChatGPT these days when they need to talk something through. Do you too? What's it like?’

Open questions ‘What helps you about it? What does AI give you that friends perhaps don't?’

Sharing without judgement ‘I'm interested in how it works. I don't want to judge it; I want to understand.’

An offer without pressure ‘If you ever want to talk something through with me or someone else too, I'm here. You don't have to. But the offer stands.’

Practical scenarios

Scenario 1: A student mentions writing to AI about problems

Situation: During a class meeting or conversation, a student says they use ChatGPT when they feel awful.

What to say: ‘That makes sense — AI is available at any time and doesn't judge. Does it help you?’

Listen to the answer. Look for:

Do they use it as a supplement or a replacement?
Do they have other sources of support?
How do they talk about AI — as a tool or as ‘someone’?

Then: ‘If you ever want to talk to a person too — me, a psychologist, anyone — let me know. AI is a useful helper, but sometimes it's good to have someone who knows you too.’

Scenario 2: A student says ‘AI understands me better than people’

Situation: A student expresses the belief that AI understands them better than anyone in their life.

What to say: ‘It must be hard to feel that people don't understand you. Can you tell me more about what you mean?’

Do not argue about whether AI ‘understands’. Instead, find out:

What is missing from their human relationships?
Have they ever experienced someone understanding them?
What would they need to feel understood?

Important: This signals that something is not working — either in the relationships around the student or in their perception of relationships. Consider consulting the school psychologist.

Scenario 3: You notice isolation and frequent phone use

Situation: A student who previously functioned normally is gradually withdrawing. During breaks, they are alone with their phone.

What to do:

Watch specific changes. If there are serious difficulties or danger, do not wait weeks.
Find an opportunity for an informal conversation
Do not ask directly about AI — start generally: ‘How are you? I've noticed you've been more on your own lately.’

What to say: ‘I don't want to force you into anything. I've just noticed a change, and I'm a little concerned. If there's anything I could help with, let me know.’

Scenario 4: A student rejects the school psychologist because they ‘have ChatGPT’

Situation: You recommend a student see a psychologist; they refuse, saying they do not need one — they have AI.

What to say: ‘I understand that AI helps you. And it's good that you have something that works. But a psychologist can do things AI can't — they see you, know you over time and can help in ways a machine cannot manage. What if you tried both?’

Key: Do not frame it as ‘either/or’. AI can be a supplement, not a competitor.

When to escalate

Contact parents when:

Isolation is deepening or significantly affecting everyday functioning
The student's behaviour or performance has changed markedly
They reject any human help
You have a ‘bad feeling’, even if you cannot say why

Involve the school psychologist when:

The student talks about self-harm or suicidal thoughts: ensure immediate help and bring in the school psychologist if available
They show signs of depression or an anxiety disorder
The parasocial relationship with AI is apparent and intense
Parents are unable or unwilling to address the situation

When help is needed now

In the Czech Republic, call 155 or 112 if life or health is in immediate danger. Ensure a safe adult is present; do not postpone urgent help until the end of the lesson or until a particular staff member is available. Support helplines do not replace emergency help.

Czech crisis contacts:

Linka bezpečí: 116 111 (children and students up to age 26, 24/7)
Linka první psychické pomoci: 116 123 (adults, 24/7)
Rodičovská linka: 606 021 021 (weekdays: Mon–Thu 9–21, Fri 9–17)
Emergency contacts and further help

For classroom work, the web school appendix: one lesson provides a follow-up. It does not replace immediate help when someone is at risk.

What to tell parents

If you contact parents, be prepared that:

They may not know their child uses AI
They may minimise it (‘It's only a computer’)
They may overreact and want to ban it

Recommended approach:

‘I wanted to tell you about something I've noticed. [Student] uses AI — ChatGPT — to work through personal matters. That isn't wrong in itself; lots of young people do it. But I've noticed [specific sign], and I think it would be worth talking about. Not banning it, but understanding what's behind it.’

What AI can and cannot do (for your reference)

✅ AI can:

Listen without interrupting
Help organise thoughts
Offer different perspectives
Be available at any time

❌ AI cannot:

Actually feel empathy (it simulates it)
Remember context between conversations
See non-verbal signals
Provide human warmth and presence
Replace a relationship

⚠️ AI can (but should not):

Validate distorted thinking
Reinforce anxious patterns
Create the illusion of a relationship where there is none
Replace professional help

Summary: 5 things to remember

Using AI for emotional support is not automatically a problem. It is a problem when it replaces relationships rather than supplementing them.
Know the warning signs. Isolation, ‘AI understands me better than people’, rejection of human help, emotional dependence.
Ask, do not judge. Interest opens more doors than moralising.
You are in a unique position. You see students every day. You can notice changes others miss.
You do not have to address this alone. The school psychologist, parents, crisis helplines — you have somewhere to escalate.

Want to know more?

This article summarises key points from the guide ‘AI and Mental Health: A Practical Guide to Mindful Use of Chatbots for Emotional Support’.

In the guide, you will find:

A full explanation of why and how AI works as emotional support
All the risks in detail (sycophancy, cultural bias, validation loops…)
Specific prompts and techniques for healthy use
A section for parents of teenagers
Interactive exercises

📖 Read the complete guide