Paste one real student answer into the prompt
Models tune to examples better than to descriptions. One genuine, muddled answer teaches the AI more than three sentences about your class.
Jul 13, 2026 · 1 min read
Describing your students ("mixed ability, some EAL, common misconceptions about X") produces generic differentiation. Showing the model one real piece of student work, with every name and identifying detail stripped out, anchors it to what your class actually writes.
In practice
- Use it for feedback drafting: "Here is an anonymised student answer. Draft two sentences of feedback: one strength, one next step, in language this student will understand."
- Use it for planning: "This answer shows a typical misconception from my class. Design three practice questions that target exactly this gap."
- Strip names, and anything identifying, first, every time. UNICEF's policy guidance on AI for children (UNICEF, 2021) puts data protection among its core requirements, and a consumer chatbot is not a school data processor. When in doubt, retype the relevant fragment rather than pasting a photo or full document.
Anonymise before you paste. Student work is student data.
Quick takeaway: one real example beats three sentences of description.