ChatGPT, Gemini, Claude, and the like have long been part of everyday school life, and in many cases, they are already taking on tasks that should actually be reserved for student thinking. Students use them to solve homework, write presentations, or summarize content. In many cases, these systems deliver quickly understandable and linguistically convincing results. In classes where digital devices are available, they sometimes.
This development is taking place more rapidly than pedagogical concepts or clear rules can emerge. While new subjects and educational reforms are being discussed, practice has long since changed – often without clear direction. The crucial question is therefore no longer whether artificial intelligence is part of school, but what effects it has on learning processes – and how to deal with it responsibly.
Why many parents' skepticism is justified
The hesitation of many parents towards AI in the school context is not unfounded. It stems from concrete observations and legitimate concerns:
Tasks are completed without the underlying thought process being visible
Results appear
Children get used to immediately using external help when facing difficulties.
Independent problem-solving is practiced less often
The difference between „understood“ and „gotten delivered“ blurs.
These points do not only concern individual subjects or situations, but the fundamental understanding of learning. Learning is not just an outcome, but a process – and it is precisely this process that is under significant pressure due to the use of AI.
The central problem: When thinking is outsourced
At its core, it's not about technology, but about a shift in the learning process. In research, this is referred to as „cognitive offloading“ – the outsourcing of cognitive tasks to external systems.
Fundamentally, cognitive offloading is nothing new. Humans have always used tools like notes, calculators, or reference books. However, it becomes problematic when not just supportive functions are offloaded, but core thought processes themselves:
Analyze task requirements
Structuring thoughts
Developing solution strategies
Checking results
Formulate your own arguments
If these processes are regularly replaced by AI, a fundamental learning problem arises: the skill is not practiced because it is no longer needed.
What current studies actually show
The discussion surrounding AI is often characterized by strong opinions. This makes a look at empirical findings all the more important.
The Relationship Between AI Use and Critical Thinking: Better Results—But Weaker Learning
A study by Michael Gerlich (2025) shows a clear connection between the use of AI tools and critical thinking ability. The focus is on so-called „cognitive offloading,“ which means outsourcing thought processes to external systems.
The results indicate that the more AI is used, the more cognitive processes are outsourced, and the lower the critical thinking scores become. Younger users are particularly affected, while a higher level of education acts as a protective factor.
The central realization here is not that AI is fundamentally problematic, but that the way it is used is crucial. Those who regularly outsource thinking stop training it – and therein lies the real risk. If AI becomes a substitute for one's own thinking, it can weaken essential cognitive abilities in the long run.
Better results – but weaker learning
An experimental study by Hamsa Bastani and colleagues (University of Pennsylvania) shows a similar effect in the academic context. Students who use AI support in their learning initially achieve better results on practice assignments.
However, in tests without AI support, they sometimes perform worse than students who learned without AI. Thus, the short-term performance gain does not automatically lead to sustainable learning.
The explanation is that AI is often used as „performance support“: it helps to complete tasks correctly in the moment, but in doing so, it partially replaces one's own thought process. However, a correct result without personal understanding is not learning – it is merely a simulation of it. Therefore, it is also crucial here whether AI supports the learning process or replaces it.
OECD findings
The OECD confirms this observation in its Digital Education Outlook (2026). AI can support learning processes if it is didactically integrated. However, without a clear pedagogical structure, it leads to superficial processing rather than sustainable learning.
The central distinction is:
AI as a tool to support learning processes
AI as a Replacement for Learning Processes
Genuine learning gains arise only in the first case.
Influence of Trust in AI
A study by Microsoft Research (2025) shows that trust in AI has a decisive influence on behavior. Individuals who trust AI strongly tend to invest less of their own cognitive effort. At the same time, high trust in one's own abilities acts as a protective factor.
Applied to students, this means: those who feel insecure are particularly susceptible to handing over their thinking to AI.
Why a ban is not a solution
Given the risks described, which are certainly not fully elaborated here, a ban on AI in the school context initially seems obvious. However, this approach is neither realistic nor effective in practice. AI systems are available at all times and are also used by students outside of controlled learning environments. A ban would therefore not lead to non-use but merely shift its use to a non-transparent area where neither guidance nor reflection takes place.
Therefore, a sustainable approach to AI doesn't require avoidance, but rather the targeted development of competence. It is crucial that students learn to use these tools consciously, reflectively, and purposefully.
When AI Learning Actually Weakens
AI has a particularly negative impact on learning processes when it replaces central steps of thinking. Its use becomes problematic primarily when students receive solutions before they have attempted to solve a problem themselves, when complete texts are copied without genuine understanding, or when calculation methods and reasoning are no longer fully understood and verified.
It is equally critical when summaries are used without engaging with the original content, or when errors are corrected without understanding why they occurred. In all these cases, time is saved in the short term, but at the same time, the training of central cognitive skills is forgone.
When AI can support learning
At the same time, it is becoming apparent that AI can make a meaningful contribution to learning under certain conditions. It is crucial that it is used as a support rather than a replacement.
AI can promote learning processes when it is used to make content more understandable, provide alternative explanations, or enable targeted follow-up questions. It can also provide meaningful support during practice and repetition, especially when it provides feedback without taking over the actual thought process.
The crucial point always remains the same: active engagement must stay with the learner. AI can accompany the process, but not replace it.
The crucial question: Who thinks first?
A central approach to the responsible use of AI in learning can be reduced to a simple guiding question: Who performs the initial thinking process – the student or the AI?
A sensible process involves the student first attempting to solve a problem independently. AI is then used specifically to obtain hints or explanations only when concrete difficulties arise. These must subsequently be understood and comprehended. Crucially, the student must finally be able to explain or apply the solution path again without support.
Only when this last step succeeds can we speak of sustainable learning.
So the rule is: He who thinks first learns best.
How parents can specifically support their children when using AI
Parents bear the primary responsibility for their children—and therefore also for what and how they learn. They should not simply transfer this responsibility to the school or allow themselves to be gradually relieved of it. Schools can impart content and set frameworks, but whether a child develops thinking processes, learns perseverance, and builds understanding is largely shaped in the home environment. This responsibility cannot be delegated to the school or to digital tools.
This does not mean that parents need to master every technical task themselves. Rather, it is crucial that they consistently focus on the learning process and actively support it.
Specifically, this happens, for example, through targeted follow-up questions: What was tried first? Where did a problem arise? Why is a solution correct? Can the solution path be explained comprehensibly - even without a screen?
Such questions shift the focus from the result to the thought process. That is exactly where sustainable learning occurs.
Consequences for teaching and assessments
If AI can reliably provide answers, the mere retrieval of results loses its significance. Therefore, it becomes crucial to design learning processes in such a way that thinking becomes visible and verifiable. This is especially true for homeschooling, as learning processes are less immediately apparent there and rely more heavily on self-responsibility.
For teaching, this means a clear shift: away from tasks that primarily focus on the final result, and towards formats that put the thought process at the center. Students should not only show that they arrive at a solution, but how they get there, what alternatives they consider, and how they justify their results.
Concrete measures
Concrete measures in everyday school life can include:
Students explain their solution methods in their own words and justify each intermediate step.
Students compare their own approaches with AI responses and critically evaluate them (What is correct? What is missing? Where is it inaccurate?)
Students analyze typical errors and explain why they occur and how to avoid them.
Students first solve a task without AI and then work on it with AI – and reflect on the differences
Students transfer learned principles to new, unknown tasks (transfer).
These examples are intentionally meant to be a selection. Creativity is called for from the perspective of teachers and parents. The goal is not the perfect method, but rather to design learning formats in such a way that the student's ability to think is demonstrably promoted – particularly understanding, reasoning, transferring knowledge, and critical thinking.
Proof of performance
For evaluating performance, it will be crucial to choose formats in which the thought process becomes visible, not just the outcome. Formats in which the use of AI is only possible to a limited extent or becomes transparent are particularly effective. These primarily include oral examinations, short presentations, spontaneous explanation phases in class, and dialogical situations in which students present and justify their solution path live.
At the same time, a clear limit of written evidence becomes apparent: even if tasks are consciously designed to encourage justification, reflection, and independent thinking, AI can convincingly meet these requirements – including derivations and structured arguments. The challenge, therefore, is less about formulating tasks that are „AI-proof“ and more about reliably recognizing genuine individual achievement. Written results alone are increasingly insufficient for this purpose.
In practice, this means that written work remains useful, but must be supplemented with formats in which understanding can be directly assessed. Only in this way can the child's learning progress be demonstrably and reliably evaluated, because the underlying thought process becomes visibly traceable – and thus it becomes clear whether a student truly understands content or is merely reproducing it.
Open communication culture in dealing with AI
Regardless of the learning setting, it is crucial that adults actively seek dialogue with students and guide them in dealing with AI. Regular, open conversations about their learning journey, difficulties, and the opportunities and limitations of AI help develop a reflective usage behavior. Parents, in particular, bear the primary responsibility here; at the same time, schools are also responsible for developing appropriate coaching and support formats that systematically train students in the reflective use of AI.
The goal is not to leave students alone with the tools, but to support them in making conscious decisions: When does AI really help? When does it replace their own thinking process? Such a culture of discussion promotes independence, a sense of responsibility, and the ability to use AI critically and meaningfully.
Basic rules for the responsible use of AI
- He who thinks first learns best.
- AI can help, but you have to learn yourself.
- You only truly understand something if you can explain it yourself.
- What you can't do without AI, you can't do.
- AI is only as good as the person questioning it.
These rules create a clear framework in which AI is used meaningfully and the student's critical thinking remains central.
Conclusion: AI as a Tool – Not a Replacement for Thinking
Artificial intelligence will permanently change everyday school life. It offers new possibilities, but at the same time places fundamental demands on how learning is approached.
The crucial question is not whether AI is used, but how. If it replaces thought processes, it leads to superficial learning. If it supports thought processes, it can be a valuable tool.
In the long run, it will not be those who find answers as quickly as possible who will be successful, but those who have learned to think for themselves. Only these students will grow into capable individuals who can take responsibility.
