How teachers can become digital creators through generative AI
Artificial intelligence (AI) is no longer a future scenario – it is already profoundly changing the education sector. However, while students are increasingly using AI-supported tools such as ChatGPT, many teachers are faced with the challenge of learning how to use this technology themselves. How can generative AI (genAI) be meaningfully integrated into teaching? What skills are needed to use this technology critically and reflectively?
The new publication “From Gretel to Strudelcity: Empowering Teachers Regarding Generative AI for Enhanced AI Literacy with CollectiveGPT” by Benedikt Brünner, Sandra Schön and Martin Ebner (Graz University of Technology) [doi.org/10.3390/educsci15020206] deals with precisely these questions. The study shows how targeted AI training for teachers can be designed – and provides a concrete workshop design that has already been tested with 191 participants.
Why do teachers need AI literacy?
Generative AI offers enormous potential for teaching and learning, but it also presents challenges: How are AI-generated answers created? How does the training data set influence the results? And what are the risks in terms of manipulation and bias?
In UNESCO’s AI Competency Framework for Teachers, it is emphasized that teachers must have basic AI skills in order to support students in an increasingly AI-based learning environment. This includes, among other things:
✅ Understanding of AI technologies
✅ Critical reflection of AI results
✅ Recognizing ethics and bias in AI systems
✅ Developing teaching applications
But what form could such training take in practice? This is precisely where the concept of Brünner, Schön and Ebner comes in.
The CollectiveGPT workshop model: interactive, practical, and scalable
The authors developed an interactive workshop format specifically tailored to the needs of teachers. The basic idea is to enable teachers to experience first-hand how generative AI works – not just in theory, but through practical tasks and comparisons with human responses.
This is how the workshop works:
1️⃣ Part of the collective response: Through targeted prompts (“Hänsel and ___”, “The capital of France is ___”), participants compare their own answers with AI-generated texts.
2️⃣ Contextualization of data: Shows how AI answers change depending on how they are formulated and embedded.
3️⃣ Training: Helps participants to recognize that AI does not fill real “gaps in knowledge”, but only reproduces known patterns.
4️⃣ The Eureka Moment: A hidden manipulation trick is revealed – a systematic “prompt injection” shows how easily AI results can be influenced.
The workshop starts right here and covers all these topics from a school context. The aim is to convey questions such as “How can AI enrich teaching?”, “What challenges does it present?” and “How can students learn to critically question AI results?” in a practical way.
Results: Significantly improved AI skills
An evaluation of the workshops shows that 70% of teachers rate their AI knowledge as better after the workshop. The average grade improved by a whole school grade! The interactivity, practical relevance and comprehensible examples were particularly praised.
“Strudelcity” and the power of manipulation – a highlight of the workshop
A special highlight was the experimental prompt with “Strudelstadt”. The participants and the AI were given the task of completing a story. In one funny story, the capital of France was changed to “Strudelstadt.” Both the AI and humans adapted to this manipulated prompt and “forgot” Paris as the actual capital – an impressive example of the susceptibility of LLMs to manipulation.
Through such playful but concrete experiments, teachers developed not only a technical understanding of generative AI, but also a critical sensitivity to its risks.
Conclusion: a model for the teacher training of the future
The study not only provides a scientifically based approach, but also a directly applicable workshop design that can be used in teacher training, schools or universities. The results speak for themselves:
💡 High acceptance: Teachers appreciate the practical, interactive approach.
📈 Measurable increase in competence: Self-assessment of AI skills improves significantly.
🔄 Scalability: The workshop concept works both online and in person and is flexibly adaptable.
With this innovative approach, the authors help to equip teachers with the skills to use AI in the classroom. After all, only those who understand AI can use it sensibly and responsibly!
📖 Find out more:
🔗 Original publication: MDPI Open Access
Brünner, B., Schön, S., & Ebner, M. (2025). From Gretel to Strudelcity: Empowering Teachers Regarding Generative AI for Enhanced AI Literacy with CollectiveGPT. Education Sciences, 15(2), 206. https://doi.org/10.3390/educsci15020206
🔗 GitHub repository for the “CollectiveGPT” tool: github.com/ed-tech-at/CollectiveGPT
🔗 Blogpost about the workshop: ed-tech.at/2024/collective-gpt/
📧 Would you like to contact the author directly? The corresponding author can be reached at
