On this page you can download the discussion paper that was submitted for publication in the Journal for Media Linguistics. The blogstract summarises the submission in a comprehensible manner. You can comment on the discussion paper and the blogstract below this post. Please use your real name for this purpose. For detailed comments on the discussion paper please refer to the line numbering of the PDF.
Blogstract of
Writing a Fairy Tale with a Little Help of ChatGPT — Experiences of Fourth Graders
by Alyssa Kutzner & Kirsten Schindler
“I am a primary student in fourth-grade. I want to write a fairy tale. Tell me 15 figures I could use”. In this or another way student writers address ChatGPT while writing their text – a creative fairy tale of their own – collaboratively. This study explores how young learners engage with generative language models like ChatGPT– not just by consuming their output, but by learning to talk to them. Through a detailed analysis of 92 prompts and their resulting texts, we take a closer look at what we call “prompt literacy”: the emerging skill of knowing how to phrase questions and requests in ways that guide the AI to produce useful, creative, or context-sensitive responses.
In our article, we propose a methodological approach for analysing chat protocols with a special focus on the input the writers produce. We also suggest understanding writing prompts as a new writing strategy (i.e. “prompt literacy”) writers have to establish. To support this writing strategy, insights into AI-based writing processes are crucial. Chat procedures are a possible path in doing so. Rather than treating prompts as incidental, this research systematically categorises the structures students used. We distinguish between imperative requests (“Tell me five names”), yes-no questions (“Do you have an idea for a fairy tale?”), and wh-questions (“What rhymes with Enis?”) – all of which reflect different communicative intentions. To make sense of these, we developed a simple rule-based schema that encodes each prompt in functional components like [operator], [amount], [recipient], [object], or [condition]. This allows us to uncover how children structure their requests, quantify expectations, and specify features – all while playing with language in surprisingly purposeful ways. Mostly, the students use the chatbot as a source of inspiration, information or revision. Some of their requests, such as “give me 599999 titles for my fairy tale”, are less productive but in most cases the writers use elaborate requests or questions for the non-human addressee.
But the story doesn’t end with prompts. We also analysed how students used (or chose not to use) the content they received in return. Some groups adopted AI-generated names or titles word-for-word. Others modified the suggestions to better fit their own narrative ideas. Still others ignored the model’s output altogether, treating it as a kind of silent collaborator: useful for inspiration, but not for actual content.
So what does this tell us? First, that prompting is more than asking questions – it’s a form of design. Students are learning to shape their own ideas through dialogue with AI. Second, even very young users are capable of using generative tools critically and creatively. And third, AI can function not only as a writer, but as a co-author, companion, or even invisible muse, depending on how students decide to work with it. As educators and designers of future learning environments, we should be paying attention. Prompt literacy is a real, teachable skill and one that will become increasingly relevant as generative tools continue to evolve. This study offers a first step toward understanding how children learn to speak to machines and how those conversations, in turn, help them tell stories.