Give it the context a colleague would need. An AI does far better with “I teach second-year pharmacology to 180 students, mostly nursing, and they struggle to apply clearance to dosing decisions” than with “help me teach clearance”. Say who your students are, what they already know, what goes wrong, and what you’re trying to achieve. The chat type, discipline and level do part of this for you; your first message should do the rest.
Ask for something specific, then iterate. “Give me three short in-class activities on renal clearance, each under ten minutes, that need no equipment” is far more useful than “ideas for teaching clearance”. Treat the first reply as a draft. Push back, ask for alternatives, say what doesn’t suit your students, and ask it to explain its reasoning. The value of a chat over a single prompt is the back and forth.
Use it to think, not just to produce. Some of the best uses don’t create anything: talking through why a tutorial fell flat, testing whether an assessment really measures the outcome it claims to, or asking it to argue against your plan. Ask it to play devil’s advocate or to list what could go wrong. A good critical friend makes your thinking sharper, and the decisions stay yours.
Rehearse the conversations you dread. Telling a student they’ve failed a placement, meeting about suspected misconduct, or responding to a mark dispute are hard to practise any other way. Set the scenario realistically, including how the person feels and what they’re not saying. Try different openings, then ask for feedback on what worked. Run the same scenario again with the student taking it differently, because real conversations rarely go to plan.
Calibrate with your own criteria, not the AI’s. For marking calibration and feedback, paste in your rubric or criteria and a de-identified sample, and ask the AI to explain how each criterion applies. Use it to surface inconsistencies and to put your own judgement into words, never to decide a grade. The mark and the feedback a student receives must be yours, in line with your institution’s policy on AI in assessment.
Check everything that matters. AI replies can be fluent, confident and wrong, especially about references, statistics, recent changes, legislation and clinical detail. Check facts against an authoritative source before you teach them, and never cite a reference you haven’t found and read yourself. When you chat about a PDF, ask the AI to point to where in the document it found something, then look.
Keep personal information out. Never include names, student numbers, health details or anything else that identifies a student, patient or colleague. Change details in scenarios and de-identify any student work you upload. The chat is for your thinking and your materials, not for records about people.
Turn good conversations into material. When a chat has produced a clear plan, a set of examples or a sharp explanation, don’t let it sit in your chat history. Send it to the Lesson Planner, the MCQ Creator, the Mind Map Creator or the Case Study Creator, or download it and build on it. The chat is where ideas get worked out; your teaching materials are where they get used.