For teachers

A practical guide for day-to-day school with AI — factual, evidence-based, without ban-reflection and without technology-euphory.

Schools are under pressure today, which nobody knew ten years ago. On the one hand, they are supposed to prepare children for a world where AI tools are as natural as today's calculator. On the other hand, they should continue to impart independent thinking, writing skills, critical source work and linguistic precision — that is precisely the skills that deteriorate the fastest through unreflected AI use.

This balancing act cannot be solved by prohibitions. Prohibitions are circumvented — by pupils, by parents, by smartphones that are always there. Prohibitions also prevent the school from making AI visible and verifiable at all. But the balancing act cannot be solved by unneeded enthusiasm — not every AI application that brings time efficiency is educationally sensible. What it takes is a conscious educational attitude.

This page attempts to translate such an attitude into practical notes. It does not replace any cantonal directives and no discussion in the college. It is intended as a personal working document for the teacher — as what you can read on Sunday evening when the preparation of the week is before you.

Four key questions before each use of AI in a classroom situation:

Does KI promote learning or just the production of results?
Are sources, biases and boundaries explicitly addressed?
What data is flowing — and do students know that this is happening?
What linguistic and mental effects are already visible in my class's work?

Understanding the pedagogical double function

Each task you give has two effects: the result (the finished essay, the solved math example, the unit) — and the learning process that the child undergoes during his work. In traditional tasks both come together. They're falling apart with AI.

A child can deliver a predictable result today with two clicks without having passed the learning process behind it. From the point of view, everything is fine. From the point of view of child development a decisive exercise unit has failed. To make this double function explicit — to itself, to class, to parents — is the most important first reaction to the AI question.

Where AI makes sense in teaching

There are clear fields where AI supports learning instead of replacing it:

as a search entry, followed by source examination on original texts
to explain concepts in alternative formulation (for children who do not have a certain style of explanation)
as a translation hypothesis that the child then critically examines himself
as a writing coach with feedback on structure and argumentation line — not as a ghostwriter
to create training material according to the teaching staff
as a Sparring partner for democratic discussions in which the class asks the AI

Where AI hurts in class

There are also clear areas where AI use threatens learning:

Homework whose actual purpose is the process of self-shaping
Reports and presentations in which the AI takes over completely — even if the result looks correct
Answers to personal questions without notice that this conversation is documented
Research on controversial topics without subsequent discussion of the source
Supporting emotional difficulties — people are responsible for this, not language models
Assessment situations in which the instructor cannot check what comes from the child and what comes from the machine

Classroom rules that work in practice

From the experience of schools that have been consciously integrating AI for two or three years, some rules can be distilled that bear in everyday life:

Transparency requirement: Any use of AI is identified — for what, to what extent. Not as a penal measure, but as a standard practice such as references.
Same tool for all: If AI is allowed, it is allowed for the whole class. Otherwise, inequalities are created which weigh more severely than the educational effect.
Orally before writing: topics that are important are discussed orally first. Only then shall the written processing, with or without AI, begin.
AI-free spaces: At least one compartment per week remains explicitly AI-free. There, what else is lost is practiced.
Reflection: Every work created with AI is complemented by a short reflection text: What was my question? How did I check the AI response? What have I changed?

Special considerations

The effect of AI is not the same in all subjects. Here is a first orientation — it does not replace a specialist depression, but it helps to identify the most important stumbling traps.

Languages (German, English, French, Italian): AI is the most dangerous here. Language is personality, and language is learned by mimicry. Those who use KI in writing lessons take their style. Keep writing consistently AI-free or request AI-track-free versions where the child listens to himself.
History and Political Education: AI is dramatically distorted here. She mainly knows English-language sources and takes over their perspective. The Swiss, the continental European, the non-western view almost always lacks. Use this as a teaching opportunity: let the class compare the AI response with schoolbook and original sources.
Mathematics: AI calculates reliable — but it calculates, it does not understand. Anyone who uses AI for task solutions learns mathematics like recipe code: correct result, no understanding. Reserve at least half of the practice time for AI-free computing.
Natural sciences: Here AI can help greatly in the concept explanation, but it regularly hallucinates studies, numbers and author names. Always request an original source as proof.
Art, music, literature: AI output is statistical — it produces average. Anyone who thinks art only about AI never learns to develop their own taste. Keep these compartments sacred as a training field for personal expression.
Computer Science and Programming: AI is now an integral part of software development. But learning programming is first without being able to debugging AI. First two years: no AI assistant. After that: consciously dosed use.

Concrete scenarios from everyday life

Scenario 1 — The too perfect attachment. A student submits an essay that has a striking effect on content and language. You suspect AI use. What to do? Avoid the detector trap: AI detectors are statistical estimators with high error rate. Instead, talk to the student: “Tell me in three minutes what it is about in your text.” Who really wrote the text himself can. Whoever has it from the AI fails.

Scenario 2 — The ratless pupil. A student refuses to do homework without doing AI. “I can no longer.” This is a warning signal, not an occasion for disciplinary action. Take ten minutes for a single interview. Offer a step-by-step reconciliation: first the search, then a separate sketch, then KI only for structuring.

Scenario 3 — The desperate child. An eight-leafer tells her that she always talks to ChatGPT because nobody else listens. This does not belong to everyday teaching, but to a professional conversation. Add the school social work, inform the parents, contact Pro Juventute 147. Don't leave it with a nice word.

Scenario 4 — The parent question. A couple of parents asks in the quarterly conversation: “Should our child use ChatGPT or not?” Answer honestly: 'It will do anyway. The question is whether we accompany it or not.” Reference to the parent page of this platform — it is written exactly for such conversations.

When it's time to hit an alarm

Pay attention to these signals in individual children:

The language of the child is clearly approaching the AI language — balanced, formulated, without corners
The child can no longer start a written task without AI
Personal topics are first discussed with the AI, then only — if at all — with parents, siblings, teachers
The level of attention in teaching is significantly reduced
Frustration tolerance in difficult tasks collapses: the child gives up after three minutes because “the AI knows faster”
The child defends AI responses against the teacher’s correction — the machine has become more authoritarian than the human reference person

Several of these signals at the same time are an occasion for a longer conversation — with the child, with the parents, possibly with the school psychology.

Parent communication

Parents react very differently to the AI theme: some are euphoric, other panic, many insecure. A good parent communication takes all three groups seriously without confirming them.

Three points that help the AI in each parental conversation: first, the double function (result vs. learning process) explicitly name. Secondly, suggest concrete house rules — parents want instructions, not theories. Thirdly, refer to the parent page of this platform so that they can continue the conversation at home.

Notification and handling of suspects

AI recognition is statistically unreliable today. Detector tools provide high false-positive rates, especially for multilingual students and for linguistically strong children. Don't rely on detectors as evidence.

The oral examination is more reliable: a brief conversation about your own text almost always reveals AI use. Set evaluation formats so that performance remains visible — sketches, preliminary discussions, source information, reflection parts. A pure final product evaluation invites you to use AI.

If you have suspicions: go into the conversation without prejudice. Some children do not know where their work ends and the AI begins. This is diagnostically more important than the note.

Teacher and the speed of change

One last comment that is rarely expressed: the speed with which AI changes is also burdensome for the teacher. What was considered safe in spring is overtaken in autumn. No one can track every day all tools, all models, all updates. You don't have to.

Concentrate on what does not change: the basic skills that you convey (thinking, language, source work, personality). The tools will be different, the skills remain the same. Whoever keeps an eye on it comes with less breathless through change.