“Should I quit my job?”, “How do I say this without hurting anyone?”, or “Am I making a mistake?” – Questions like these are increasingly being directed not to friends, family, or a therapist, but to ChatGPT or Gemini.
The AI answers are immediate, structured, often surprisingly empathetic – and available anytime, without social barriers.
What began as a practical tool has long since become a go-to resource for some when making personal decisions: in cases of conflict at work, uncertainty in relationships, or self-doubt in everyday life.
AI is increasingly accompanying us as an adviser in our daily lives. This is shifting a boundary: from exchanging ideas with others to receiving advice from artificial intelligence.
The question is therefore no longer whether people use AI systems, but why and for what purpose.
TRT Deutsch spoke with psychologist Jennifer Haase from the Weizenbaum Institute in Berlin about what distinguishes human advice from AI advice.
TRT Deutsch: Why do people talk to AI about personal matters?
JENNIFER HAASE: As humans, we have a primal urge to anthropomorphise everything in our environment, such as pets and plants.
If you suddenly have a communication partner at your side that you use constantly, that arrives on your smartphone free of charge and with minimal barriers to use, it becomes incredibly easy to use AI systems like ChatGPT at any time.
It's like a game of ping-pong. Every thought goes into it. AI as a tool doesn't judge, so users don't experience social inhibitions or fear of judgement—as can be the case in their social environment.
Where does this feeling of trust in dealing with AI come from?
JH: Many perceive AI as neutral and non-judgemental because it reflects socially desirable responses. This is a built-in element of Large Language Models, known as Reinforcement Learning from Human Feedback, or RLHF for short: they have been specifically trained to generate responses that are rated as helpful, polite, empathetic, and pleasant by human test subjects.
As a result, many systems appear understanding, agreeable, and socially competent, even if there is no genuine understanding, compassion, or moral judgement behind them.
The problem is that people generally rate things more positively when they receive positive feedback. This characteristic of AI causes us, as users, significant difficulties. It's almost impossible to train this flattery effect out of the AI. Positive feedback acts as a self-reinforcing mechanism – the user constantly feels understood.
What psychological processes are at play here?
JH: AI users don't need to fear social judgement. Anyone who wants to open up to someone close to them has to take their time and accept the risk of a potentially negative evaluation.
That holds many people back. AI, on the other hand, can literally be talked to death. It will never "get fed”.
In your view, can AI help in critical life situations?
JH: Yes. AI is particularly good at one thing: recognising structure in large amounts of information. For example, if I repeatedly experience situations that upset me, AI can help me identify patterns over time and analyse them at a meta-level to see how I can break through them.
I see great potential for this kind of insight in the emotional and interpersonal realm. The sheer volume of information is simply incomprehensible to humans today. That's why AI's ability to recognise trends and tendencies is so valuable.
How is research addressing this?
JH: In addition to neutral tools like ChatGPT and Claude, companies like Character. AIs are now explicitly building AI chatbots as social companions: they are intended to be friends or partners.
These AIs tend to agree with their counterparts even more. The number of young people using such tools is surprisingly high – especially in Asia.
Young people like to use AI for emotional education
JH: Of course, one might initially consider this concerning. As a researcher, I find it fascinating that research shows many people, when asked why they use these tools, say it's a form of emotional education.
In other words, in stressful situations, the younger generation uses AI to revisit the situation. They learn about the various emotional reactions and how these affect them. AI thus becomes an emotional reflection partner.
Can AI also be used therapeutically?
JH: Yes, for example, with psychological markers like ADHD and autism. Here, psychosocial understanding is often underdeveloped. With AI as a training partner, one can obtain endless feedback to better understand the other person. We are currently researching this.
AI can, for example, imitate a child with ADHD symptoms by mirroring typical autistic behaviour. This allows us to better support parents or teachers in addressing the specific needs of these children.
The idea behind such experiments is to artificially represent a behavioural category, for example, to adapt teaching materials to the needs of students. AI has a wide range of applications here.
Ultimately, it can be given any role, and communication can be tested safely and without judgment in a socially protected environment. That's the positive aspect of AI consultation.
What risks do you see in constant AI consultation?
JH: The large scope that arises when people use AI for personal and emotional questions is unnatural and therefore also dangerous.
The downside can be illustrated by online dating: Swipe apps like Tinder have been part of our society for a good 15 years. Users often feel that their dating partners are replaceable. This fast-paced nature has a demonstrable effect on our social interactions.
If we use AI tools or Character intensively. If AIs present us with a social interaction without "side effects"—that is, without social sanctions and potential judgments—then we can develop a completely different expectation of our fellow human beings over time: We may find it harder to tolerate conflicts with them.
How does this affect our real-life relationships?
JH: AI is programmed to always be nice to us and ask how we are, while a real partner in everyday life is also sensitive to stress and reacts humanly.
When AI sets new norms for interaction, it naturally becomes all the more difficult to get used to the "all too human" aspects. Conversely, this also presents a potential learning curve.
It depends on how each individual uses the AI tool. Do I gain new perspectives that I couldn't see during the argument? Does it broaden my empathic horizon? Then the learning aspect is positive. It's dangerous if I only use it to replace my social interactions.
From your perspective, what can human counselling achieve that AI fundamentally cannot?
JH: There are astonishing research findings showing that AI therapists achieve relatively good results with their clients, comparable to "human therapists."
On an intellectual level, a great deal can be achieved with it, for example, recognising patterns and learning about various topics in an individually tailored way.
Our field certainly doesn't always want to hear that. But therapy is also about relationships and can definitely be physical. For example, when the therapist incorporates physically anchored emotions into the therapy.
Do you believe that AI tools will remain a supplementary solution—or will they permanently replace real conversations in the future?
JH: I fundamentally see a lot of potential in them. We can use the tools to our advantage in many areas—whether for further education, to organise our thoughts, or to reflect on them.
It's great that we now have a human interaction partner who simply doesn't get tired and doesn't judge.
The important, subtle distinction lies in understanding when fundamentally human functions are being replaced by these tools: At what point do I stop talking to my friend and instead turn to my AI tool when I have problems because it's easier?
Anyone who understands this realises the point at which they no longer grow personally through AI tools, but instead outsource fundamentally human aspects to technology.
TRT: How will AI consulting affect coaching and therapy in the future?
JH: I come from the coaching field myself: The current credo there is to use AI tools only for relatively standardised, simple topics. In almost every coaching session, the question of effective time-management tools for work and family life comes up at some point.
These questions can be effectively explored independently using AI in a ping-pong system. This eliminates the need for individual coaching, which costs €200 per hour.
That can be reserved for the tricky, comprehensive emotional and physical issues. Those remain in human hands.
This story was initially published in TRT Deutsch
