The New Reality: ChatGPT in Swiss Classrooms
It is no longer a secret: artificial intelligence has arrived in Swiss schools. From primary school to Gymnasium, students use tools like ChatGPT, Claude, and Gemini — for homework, essays, seminar papers, and even Matura theses. According to an NZZ investigation, Gymnasium students openly admit to using AI chatbots to outsmart their teachers. One student puts it honestly: "You become lazier."
At the same time, teachers face a dilemma. They are supposed to integrate AI meaningfully into lessons — but also ensure that student work is independently written. The problem: only 9 percent of teachers say they can reliably identify AI-generated texts, according to a Bitkom study. Another 32 percent say they "somewhat" can. That means: more than half of all teachers are uncertain about this question.
What Is Happening Across the Cantons?
Basel-Stadt: Matura Thesis Gets Restructured
Basel-Stadt was the first canton to take concrete action. Starting in the 2025/26 school year, new rules apply to the Matura thesis:
- Oral examination weighted more heavily: The oral exam now counts for half the grade (previously one-third). In an in-depth subject discussion, students must demonstrate genuine understanding of the material.
- Process monitoring intensified: The development process of the thesis is more closely supervised. Students must submit a declaration of independence.
- AI use allowed but must be declared: AI may be used as a tool but must be clearly cited as a source.
- Task formats adapted: Questions and assignments are redesigned so that AI tools cannot simply complete the entire work.
The catalyst for these changes: initial fraud cases involving ChatGPT at Basel schools.
Canton of Zurich: Guidelines Instead of Bans
The Zurich Office for Public Schools chose a different path. Instead of rigid regulations, the canton developed five guiding principles that serve as a foundation for schools:
- Understand and use AI as a creative tool
- Promote understanding of AI technology
- Ensure data protection
- Strengthen critical thinking in dealing with AI
- Personal responsibility of teachers in integration
Responsibility thus lies with individual schools and teachers — an approach that offers flexibility but can also create uncertainty.
EDK: Cantons Are Responsible
The Swiss Conference of Cantonal Ministers of Education (EDK) has remained restrained. The ChatGPT matter falls within the jurisdiction of the cantons, according to the official position. Whether the topic might later also need to be addressed at the EDK level is "currently open." This means: no uniform national guidelines exist for handling AI in schools.
Lehrplan 21: AI Is Not Mentioned
Artificial intelligence is not explicitly mentioned in the Swiss curriculum (Lehrplan 21). While there is a subject area called "Media and IT" where AI can be covered, this happens only at the teacher's initiative. There is no binding mandate to prepare students for dealing with AI. Given the rapid pace of development, a remarkable oversight.
The Problem: Teachers Can Barely Detect AI Texts
A text written by ChatGPT looks flawless at first glance — correct grammar, logical structure, flowing style. That is exactly what makes detection so difficult.
Typical Clues — But No Guarantee
Experienced teachers watch for certain characteristics:
- The "ß" problem: ChatGPT uses the German "ß" by default instead of the Swiss "ss." A student from Zurich who suddenly writes "daß" or "groß" stands out — but more and more students know this and deliberately correct it.
- Unusually consistent style: AI texts often have a striking consistency in sentence length and word choice that is atypical in student work.
- Missing personal touch: AI-generated texts are correct but often "soulless" — lacking individual perspectives, uncertainties, or creative leaps of thought.
- Too perfect for the level: When a student who previously struggled with essay writing suddenly submits a flawless seminar paper, that is a warning sign.
The problem: these clues are subjective. And students are becoming increasingly skilled at disguising AI texts — for example, by instructing the chatbot to use "less intelligent words" or deliberately inserting small errors.
Manual Checking Does Not Scale
With a class of 25 students, each with a 2,000-word essay, manual AI checking means: analyzing 50,000 words individually, comparing styles, verifying sources, assessing plausibility. This is simply not feasible alongside regular teaching and grading duties.
How an AI Detector Concretely Supports Schools
An AI detector does not replace the teacher's judgment — but it provides a solid foundation for assessment. Instead of relying solely on gut feeling, teachers receive a data-driven evaluation.
What a Modern AI Detector Does
- Text analysis in seconds: Instead of manual review, the detector analyzes a text in seconds for AI-typical patterns — linguistic structures, word choice patterns, sentence construction consistency.
- Sentence-by-sentence analysis: Rather than just "yes/no," a detailed analysis shows which passages likely come from AI and which are human-written. This is particularly useful for texts where students mix AI-generated sections with their own writing.
- Multilingual detection: Switzerland has four languages. A good AI detector recognizes AI-generated texts in German, French, Italian, and English — essential for use in Swiss schools.
- Data protection: For schools, data protection is paramount. Student texts are sensitive data. An AI detector that processes data on Swiss servers and is GDPR-compliant is essential for use in public schools.
Practical Example: The Matura Thesis
Imagine: a Gymnasium student submits her Matura thesis — 30 pages, carefully written. The supervising teacher has the impression that parts of the work do not match the student's usual writing style. Instead of hours of manual analysis, she uploads the text to an AI detector:
- Result: 73% AI probability in the main body, 12% in the introduction, 8% in the conclusion.
- The sentence-by-sentence analysis shows: three chapters in the middle section exhibit strong AI patterns.
- The teacher can now specifically address these chapters in the oral examination (which in Basel already accounts for half the grade) and check whether the student truly understands the content.
The AI detector does not replace pedagogical assessment — it makes it better informed.
AI Detector as Part of a "Swiss Cheese Model"
Education experts speak of the "Swiss cheese model" for academic integrity: no single measure is perfect, but multiple layers together provide reliable protection. Each layer has holes — but the holes do not overlap.
For Swiss schools, this could look like:
- Layer 1: Task design. Assignments that require personal reflection, local references, or process documentation are harder for AI to solve.
- Layer 2: Process supervision. Regular check-ins, interim submissions, and work journals document the development process.
- Layer 3: AI detector. Technical analysis provides objective data points that complement subjective assessments.
- Layer 4: Oral examination. Conversation reveals whether a student truly understood the material — exactly what Basel now weights more heavily.
- Layer 5: Declaration of independence. The formal declaration creates awareness and a legal basis.
An AI detector is one of these layers — not the only one, but an important one.
What Schools Can Do Now
The situation is clear: AI is here, it is not going away, and bans do not work. Instead, Swiss schools need a pragmatic approach:
1. Establish Clear Rules
Define at the school level how AI tools may be used in different contexts. Basel shows how it is done: AI use is permitted but must be declared.
2. Train Teachers
Teachers need professional development — not only in using AI tools, but also in recognizing AI-generated texts and using AI detectors.
3. Adapt Assessment Formats
More oral exams, more process supervision, more personal reflection in written work. Basel leads the way — other cantons will follow.
4. Deploy Technical Tools
An AI detector that runs on Swiss servers, works multilingually, and is GDPR-compliant gives teachers a tool they urgently need.
5. Have Open Conversations
Talk to your students about AI. About the possibilities, but also about the limits. About the question of what independent thinking means — and why it is more important than ever.
Conclusion: Fairness Through Transparency
The question is not whether Swiss schools must engage with AI — but how. Cantons like Basel and Zurich show different paths. Uniform national guidelines are still missing.
In this transitional phase, teachers need support. An AI detector is not a silver bullet — but it is an important piece of the puzzle. It helps teachers make informed decisions. It protects students who do their work honestly. And it creates the foundation for what ultimately matters: fairness.
Because the goal is not surveillance. It is justice. Students who do their own work deserve to know that everyone is held to the same standard.
Sources
- SRF: "Maturaarbeit: Wegen Textrobotern setzt Basel mehr aufs Mündliche," 2024.
- NZZ: "So tricksen Gymi-Schüler ihre Lehrpersonen mit KI-Chatbots aus," 2025.
- Canton of Zurich, Office for Public Schools: "Künstliche Intelligenz in der Volksschule," 2025.
- Bitkom Study: "Teachers and AI Detection," 2024.
- EDK: Statement on cantonal jurisdiction, 2024.
- Lehrplan 21: Subject area "Media and IT."
- Federal Council: Mandate for consultation draft on AI regulation, February 2025.
- PHBern: "AI in the Classroom: China Pushes Ahead, Switzerland Hesitates," 2025.