AI prompts for lesson planning

Written for teachers. A week of plans in your own format from one prompt and a topic list — objectives, activities, and the differentiated version.

• 01 Ten sections, one for each planning job that comes back every week
• 02 30 fill-in prompts, brackets left open for your grade, your standard, your materials
• 03 30 follow-up steps — what to send next, in the same chat, to get a first draft to a plan you would submit
• 04 30 failure modes, each with the prompt that repairs it
• 05 An 84-page manual, a printable quick-reference page, and a plain-text file every prompt copies cleanly from

The first prompt in the kit does not ask for “a lesson plan on fractions.” It asks for a week. You paste your unit outline or the standard in its own wording, say how many periods you have and how long they run, and list the materials you actually have in the room. What comes back names one throughline the week builds toward, then gives each day an objective, what students do, and how you check it — with every day using something from the day before.

That first version is usually most of the way there. The follow-up chain is the rest. One step rebalances the load when the early days are thin and the last two are overloaded, and tells you what moved. One adds a daily check you can read in under a minute, plus what a wrong answer on it means for tomorrow. One rebuilds the week when you lose a period to an assembly, naming which day absorbs the cut and what must not be cut.

Ten sections work the same way: the week, the single lesson from a standard, objectives and success criteria, hooks, exit tickets, a differentiation pass on one lesson, materials and slides generated from the plan you already have, the sub-friendly conversion, and the re-plan for when today ran long.

Models fail at lesson planning in a small number of predictable ways. The kit names them before you meet them, and gives you the wording that fixes each one.

The one to watch is invented standards codes. Ask for a plan aligned to a code and a model will confidently produce a plan for a standard that does not say what you think it says. The rule throughout the kit is to paste the standard’s text and never trust a cited code — and the repair prompt makes the model quote the exact phrase each part of the lesson addresses.

The rest are familiar once named. Five days that never meet, so nothing builds on anything. Every day the same shape. A month of unit compressed into one week. Ninety minutes of activity timed for a forty-five minute period. Five objectives where one measurable objective belongs. A model that rewrites the whole plan when you asked it to change one line. Each has a repair you send in the same chat, and the manual says when to reach for which.

The free starter card carries the three rules and ten cross-cutting prompts at principle level — enough to get a usable first draft out of any model. It does not carry Kit 01’s follow-up chains, its failure modes and repairs, or the quick-reference page. That line is deliberate and it does not move.

Every prompt uses placeholders — [GRADE], [SUBJECT], [STANDARD], [STUDENT] — by construction. No student name, grade, or plan content goes into a consumer AI tool. The model drafts; you decide what a student sees.

The prompts are written to be model-agnostic: plain instructions, no tool-specific syntax, so they work in ChatGPT, Claude and Gemini, and alongside a tool like MagicSchool rather than against it. The runs behind this kit were against Claude. A dated model-notes page inside the manual says exactly that, and flags the few places where models diverge enough to need different wording.

Plate 1 · written and rendered · not yet listed