How to Make AI Lesson Plans Fit Your Real Class

Why the first plan you got back was useless

The most common complaint teachers make about AI is this one: the lesson plans are generic and do not reflect my real classroom. It is a fair complaint and it is entirely correct. It is also completely predictable once you understand what happened.

When you type "write me a lesson plan on fractions for Year 5", the tool has almost nothing to work with. It does not know how long your lesson is, what your class did last week, that seven of them still cannot find a common denominator, that four are working two years below age expectation, that you have no working visualiser, or that the last twenty minutes of Thursday afternoon is not a time for independent investigation. So it produces the average of every fractions lesson ever written on the internet. Starter, main, plenary, three differentiated tasks, a plenary question. It reads like a plan and teaches like nothing.

The tool is not failing to understand your class. It was never told anything about your class.

In plain English

Context:: Everything you tell the tool about your situation before asking for the thing. The single biggest lever on output quality.
Constraint:: A hard limit you impose, such as fifty minutes, no printing, no group work. Constraints make output specific.
Misconception:: A wrong idea students reliably arrive with. Naming yours is what turns a generic plan into a useful one.
Iteration:: Going back and refining rather than accepting the first output. Almost always where the value appears.

The fix is boring and it works

Supply the constraints you carry in your head without noticing. There are six that do most of the work.

Year group and the actual ability spread. Not "mixed ability", which means nothing. Try "six working well above, about eighteen roughly at expectation, five who need the reading load reduced, two with an adult supporting".

Lesson length and where it sits in the day. A fifty minute period after lunch is a different lesson from a ninety minute morning block.

What they did last lesson, and what comes next. This is what stops the plan floating free of your sequence.

The sticking point. The thing this class specifically keeps getting wrong. This is the highest value sentence in the whole prompt.

Your equipment and room. No printer, no devices, fixed rows, an interactive board that works about half the time. Say so.

What you want them to leave able to do. One sentence, in student-facing language.

That is five minutes of typing, and it is the difference between something you bin and something you actually teach.

Prompt you can copy: the full worked lesson plan prompt

Plan one lesson for me. Ask me nothing, work with what is below.
CLASS
Year group: Year 8
Subject and topic: Science, forces and motion, second lesson of six
Size and spread: 29 students. About 5 confident, 17 secure,
5 who find the reading heavy, 2 supported by a teaching assistant.
Behaviour note: they lose focus fast in silence, they work well in pairs.
LOGISTICS
Length: 55 minutes, straight after lunch
Room: fixed benches, no devices, one whiteboard, no printing available
today so anything written must go on the board or be dictated.
WHERE WE ARE
Last lesson: they identified balanced and unbalanced forces from diagrams.
Most got it. Next lesson: calculating resultant force.
Persistent sticking point: they treat a heavier object as automatically
faster, and they use the words force and speed interchangeably.
WHAT I WANT
By the end, every student can say in their own words what an
unbalanced force does to motion, and use the words force and speed
correctly in one sentence.
OUTPUT
Give me a timed plan, minute by minute. Include:

a 5 minute retrieval starter on last lesson
one activity that directly confronts the heavier equals faster idea
exactly what I say to introduce the key idea, as a script I can adapt
3 questions I can ask to check understanding mid lesson
2 things to do if they get it faster than expected
2 things to do if they are lost after 20 minutes
an exit ticket of one question
Do not include learning objectives written in teacher language.
Do not suggest anything that needs printing or devices.
Flag anything you are unsure about rather than inventing it.

Checkpoint

Generic in, generic out. The plan gets specific when you supply the constraints you already carry in your head: year group, spread, length, last lesson, equipment, and the thing they keep getting wrong.

❌ Weak prompt

Prompt

Write a lesson plan on fractions for Year 5.

Output

A tidy plan with a starter, three differentiated tasks, a plenary, and objectives written in curriculum language. Nothing in it is wrong. Nothing in it is about your class.

The tool averaged every fractions lesson on the internet, because that was all it had. You will spend longer adapting this than you would have spent planning from scratch.

✅ Good prompt

Prompt

Plan 45 minutes on equivalent fractions for my Year 5 class. 28 students, 9 still counting on their fingers for times tables, 4 above expectation. No devices. Last lesson they shaded fraction walls and most were secure. The sticking point is that they think a bigger denominator means a bigger fraction. Give me a timed plan, a script for the key explanation, three checking questions, and an exit ticket.

Output

A plan that opens by directly confronting the bigger denominator idea, uses the fraction walls they already know, and gives you words to say.

Same tool, same minute of work. The difference is entirely the six constraints, and the sticking point sentence did most of it.

Then argue with it

Never take the first output. The second most useful habit after supplying context is pushing back once.

Prompt you can copy: make it fit better

That plan does not fit yet. Fix these:

The starter takes 10 minutes and I only have 5. Cut it.
The main activity assumes they can already [ ]. They cannot. Replace it.
It is too quiet for 20 minutes. This class needs talk before writing.
Rewrite the explanation script at a lower reading age but do not
make it babyish. They are 13 and they notice.
Keep everything else. Show me only the parts you changed.

Save one good prompt per class as a template with the class details already filled in. Next time you change three lines instead of thirty. This is the single biggest time saver in the whole course.

Check it before you teach it

Two failure modes are worth knowing about. The first is invented content: a quotation, a date, a historical figure, a curriculum reference or a source that does not exist. Anything factual in the plan is your responsibility, and the tool will state it with total confidence either way.

The second is quieter. The plan will sometimes be pedagogically fine but wrong for your students in a way it could not have known, because you did not tell it. That is not a fault in the tool, it is a missing constraint. Add it and ask again.

Prompt you can copy: differentiate a worksheet you already have

Here is a worksheet I already use. Do not change the questions
or the difficulty of the maths.
Produce three versions:

Same questions, reading load reduced, shorter sentences, key words
defined at the top
Same questions with a worked example at the start of each section
Same questions plus two extension questions that go deeper rather
than just adding more of the same
Keep my question numbering identical across all three so the class
can work from the same board instructions.
Worksheet:
[paste it]

A plan is a starting point, not a script. You will change it in the room, because you can see thirty faces and the tool cannot see any of them. That is not the plan failing. That is teaching.

📝 Quiz

Question 1 of 4

Why do AI lesson plans come out generic?