Artificial Intelligence (AI) has made great strides in recent years and is no longer just a topic for IT specialists or futurists. Large language models such as ChatGPT have moved into the spotlight of public debate – including in schools. More and more teachers are asking themselves:
- How do such AI models actually work?
- What can they do – and what not?
- How can I use them meaningfully in class or lesson preparation?
This blog post provides a solid yet accessible introduction to the technical foundations of large language models (LLMs) and shows practical applications for everyday school life. The goal is to give teachers a well-founded orientation – without technical jargon, but with scientific accuracy.
Part 1: What are Large Language Models (LLMs) like ChatGPT?
Definition and Purpose
A "Large Language Model" (LLM) is a computer program trained to understand and generate language. It can analyze, interpret, expand, or even create entirely new texts – often in a style that seems surprisingly human.
Well-known examples include:
- ChatGPT (by OpenAI)
- Claude (by Anthropic)
- Gemini (by Google)
- Mistral, LLaMA, BLOOM (open-source models)
The uniqueness of these models lies not only in their ability to "understand" text but also in their vast knowledge base – trained on hundreds of billions of words from the internet, books, articles, and forums.
Technical Basics – Simply Explained
LLMs are based on a technique called "Transformer architecture." Here’s a simplified overview:
1. Tokenization
The text is broken down into small units – called tokens. These can be words, syllables, or even parts of words.
2. Training on Large Text Datasets
The model reads massive amounts of text. It learns statistical relationships: which word is likely to follow another?
3. Predicting the Next Token
The core mechanism: the model predicts which word or token should come next – based on the context so far.
4. No Real Understanding, Just Pattern Recognition
The model does not "understand" text like a human. Instead, it recognizes patterns, styles, and structures in language and imitates them.
A Comparison with Human Language
Human language is highly complex: grammar, context, irony, cultural references – all must be considered. LLMs are not "thinking" machines, but they are excellent at imitating linguistic patterns.
An Analogy:
ChatGPT is like an extremely well-trained parrot with a perfect memory – it knows how language works, but it has no consciousness and does not truly understand content.
Part 2: What Can ChatGPT Do – and What Not?
Strengths of ChatGPT
- Fast text creation: worksheets, lesson plans, explanations, summaries
- Varied writing styles: factual, humorous, age-appropriate, poetic…
- Multilingual abilities: translations and foreign language content (e.g., French, English)
- Idea generator: brainstorming for projects, topics, discussions
Limitations and Risks
- No fact-checking: ChatGPT can produce incorrect or outdated information ("hallucinations")
- Lack of world understanding: it does not "know" if something is moral, ethical, or historically correct
- Data protection: sensitive student data must not be entered into public AI models
- Technology dependence: risk of relying too heavily on automated tools
Part 3: LLMs in the Classroom – Applications for Teachers
1. Lesson Planning and Preparation
Example use:
- Creating differentiated worksheets
- Developing learning objectives, competency grids, or assessment criteria
- Formulating exam tasks or quiz questions
Example prompt:
"Create a worksheet on photosynthesis for 7th grade with three different difficulty levels."
Advantage: saves time, variety of ideas
Note: always review and adapt critically
2. Subject-Specific Support
LLMs can act as learning companions – for example by:
- Writing texts (language arts, foreign languages)
- Explaining terms in simple language
- Simulating roleplays (e.g., English dialogues)
- Interpreting literary works
Example classroom prompt:
"Explain Goethe’s poem 'Der Erlkönig' for 8th graders in simple language."
Advantage: student-friendly explanations, alternative perspectives
Note: literary interpretation is subjective – AI provides only one perspective
3. Differentiation and Individualization
In heterogeneous groups, AI can help create tailored content:
- Simplified texts for students with weaker language skills
- Extended content for advanced students
- Exercises in different formats (cloze texts, multiple choice, creative tasks)
Example prompt:
"Summarize fractions in simple language and provide three beginner exercises."
4. Promoting Media Literacy
Critical use of AI is part of digital education. Possible lesson ideas:
- Students analyze and fact-check ChatGPT responses
- Discussion: "Should students use AI for homework?"
- Writing workshop: "Human vs. Machine – Who writes better?"
This way, students learn to use AI responsibly and reflectively.
Part 4: Legal and Ethical Aspects
Data Protection
Important: never enter personal data (names, grades, diagnoses) into public AI tools. Many school authorities recommend caution – or provide their own GDPR-compliant tools.
This is where didactAI offers a key advantage: designed specifically for education, it meets schools’ special data protection requirements and provides a safe alternative to public AI tools.
Copyright
Texts generated by ChatGPT are usually not copyright-protected – this can be both an advantage and a drawback:
- Advantage: teachers may use them freely
- Disadvantage: no clear accountability for errors
Part 5: didactAI as a GDPR-Compliant Alternative
While public AI tools like ChatGPT have many benefits, there are also specialized solutions tailored to the needs of schools. didactAI is one such tool, combining AI strengths with educational requirements:
Advantages of didactAI over public AI tools
- Data protection compliant: designed for the education sector
- Pedagogically sound: aligned with didactic principles and curricula
- Subject-specific: adapted to different subjects and grade levels
- Structured: systematic lesson planning instead of free text generation
- Export functions: direct creation of teaching materials and lesson plans
When to Use Public AI Tools vs. didactAI?
Public AI tools (ChatGPT, Claude) are suitable for:
- Creative idea generation and brainstorming
- General text creation and editing
- Translations and foreign language content
- Discussions and roleplays
didactAI is suitable for:
- Structured lesson planning
- Creating lesson plans
- Unit planning and curriculum integration
- Professional teaching materials
- Teacher trainee lesson drafts and exam preparation
Part 6: Opportunities and Outlook – AI as Assistant, Not Replacement
LLMs are tools, not all-rounders. They can support teaching but cannot replace teachers’ educational mission.
Opportunities for Teachers
- More time for students by reducing routine tasks
- New creative impulses for lessons and projects
- Ability to support students individually
Opportunities for Students
- Learning how AI works and where its limits lie
- Developing critical thinking in digital contexts
- Motivation through interactive and modern learning methods
Part 7: Practical Tips for Getting Started
Step 1: Explore and Experiment
Start with simple tasks and experiment with different prompts. Test what works and what doesn’t.
Step 2: Integrate into Daily Work
Gradually integrate AI into your lesson preparation. Start with one class per week and expand step by step.
Step 3: Reflection and Adjustment
Regularly reflect on how AI enriches your teaching. Adjust your use accordingly.
Step 4: Peer Exchange
Share experiences and best practices with colleagues. Together you learn faster and more effectively.
Conclusion
Large language models like ChatGPT are more than hype – they have the potential to sustainably change education. But as with any new technology, understanding comes before application.
Teachers who understand how AI works and how to use it can not only make their own work easier but also prepare their students for a future where AI is part of everyday life.
For schools, a combination makes sense: use public AI tools for creative tasks and brainstorming, and specialized solutions like didactAI for structured lesson planning and material creation.
Start now with AI in teaching: Plan your first AI-supported lesson or learn more about didactAI.