AI prompts for differentiating a lesson plan three ways

Differentiating a lesson plan is the part of planning that eats a Sunday. The usual approach — three worksheets at three difficulty levels — takes hours and produces something that is not really differentiation: it is three different lessons, with the lowest group learning less.

Better differentiation keeps the learning objective identical for everyone and varies only the route to it. That is a mechanical transformation, and it is exactly the kind of thing a model does well once the objective and the specific barriers are stated. These five prompts do it — a core three-way differentiator, a variant for EAL learners, a variant for extension without acceleration, a review prompt that checks the objective survived, and a batch prompt for a unit of work.

Why does most lesson plan differentiation not work?

Because it differentiates the outcome rather than the support. If the top group analyses a source and the bottom group colours in a diagram, the second group has not accessed the curriculum with more help — they have been given a different, smaller curriculum. Do that weekly for a year and the gap widens by design.

The second problem is that differentiation is usually planned by attainment band rather than by barrier. “Lower ability” is not a barrier. Not being able to hold three steps in working memory is a barrier, and so is not knowing the four subject words the task assumes. Those need different responses, and grouping by band gives them the same one.

Third, most differentiated tasks silently change what is being assessed. A writing frame with the connectives filled in is helpful support in a history lesson and invalidates the task in an English lesson on cohesion. Whether a scaffold is legitimate depends entirely on the objective.

What should differentiation in a lesson plan actually change?

The scaffolding, the input, the amount of independent processing, and the output format — while the objective and the cognitive demand stay put.

Concretely: a partially worked example rather than an easier question. Pre-taught vocabulary rather than simplified text. A checklist that offloads sequencing rather than a reduced number of steps. A diagram or a spoken answer instead of a paragraph, where the objective is not about writing.

And crucially, a plan for removing the scaffold. Support that stays all year is not support, it is a permanent lower expectation, and it is the most common way well-intentioned differentiation entrenches a gap.

The core prompt for differentiating a lesson plan

You are differentiating a lesson plan three ways. The learning
objective must be identical for all three. You are varying the
route, never the destination.

Subject and year group: {{subject, year}}
Learning objective: {{stated as what pupils will know or be
  able to do}}
Success criteria: {{how you will know}}
What is being assessed: {{the thing that must not be
  scaffolded away}}
The core task: {{paste the task as planned}}
Lesson length and structure: {{minutes, phases}}
Prior knowledge assumed: {{what the task takes for granted}}
Subject vocabulary in the task: {{words pupils must know}}

The three groups, described by BARRIER not by band:
Group A: {{the specific barrier — e.g. working memory,
  vocabulary, reading fluency, prior gap, processing speed}}
Group B: {{barrier}}
Group C: {{barrier — including "none, needs greater depth"}}

Additional needs to account for: {{EHCP provisions, EAL stages,
  anything with a statutory requirement}}

The output format

For each of the three groups, output:

- OBJECTIVE: restate it. It must be word-for-word identical
  across all three. If you cannot keep it identical, stop and
  explain which barrier makes the objective itself
  inaccessible — that is a curriculum sequencing problem, not
  a differentiation problem.
- THE BARRIER, restated in one line.
- SCAFFOLD: what support is added, and specifically how it
  addresses that barrier.
- THE TASK: the actual task for this group, written out.
- WHAT IS NOT SCAFFOLDED: confirm the assessed element is
  still being done by the pupil.
- FIRST FIVE MINUTES: what this group does while you are with
  another group.
- REMOVAL PLAN: what has to be true for this scaffold to come
  off, and roughly when.
- CHECK: one question you can ask in ten seconds to know
  whether it is working.

Then output:
- A single set of success criteria that applies to all three.
- The vocabulary to pre-teach, per group.
- What you will need to have printed or prepared.

Rules:
- Never reduce the cognitive demand. Reduce what has to be
  held in mind at once, or what has to be decoded, or what has
  to be produced in writing.
- Never scaffold the assessed element.
- No group's task may take a different length of time.
- No colouring, word searches, or copying as a substitute
  task.
- Extension for Group C must be greater depth on the same
  objective, not the next objective.

A worked example

The brief

Subject and year: Year 8 History
Objective: pupils can explain why sources about the Peterloo
  massacre disagree
Assessed: the explanation of why accounts differ (provenance
  reasoning), not reading comprehension
Core task: read two contemporary newspaper accounts, write a
  paragraph explaining why they differ
Prior knowledge assumed: what a newspaper was, who could vote
Vocabulary: provenance, radical, magistrate, yeomanry
Group A: cannot hold the two accounts in mind while comparing
Group B: reading fluency below the source difficulty
Group C: no barrier, needs greater depth

The result

Group A — objective: pupils can explain why sources about the Peterloo massacre disagree.
Barrier: holding two accounts in mind while comparing them.
Scaffold: a two-column comparison grid with four pre-set rows (who wrote it, who paid for it, who they were writing for, what they left out). The grid offloads the holding-in-mind; it does not do the reasoning.
Not scaffolded: the pupil still explains why those differences produce disagreement. That is the assessed element and it is untouched.
Removal plan: drop to two rows next lesson, then to a blank two-column grid. Off by the end of the unit.
Check: “Which of your four rows best explains why they disagree?”

Group B — objective: identical.
Barrier: reading fluency below source difficulty.
Scaffold: the same two sources, unmodified, but heard first — read aloud once by the teacher while pupils follow — plus the four vocabulary words pre-taught with the sources visible. The text is not simplified, because simplifying it removes the period language that is evidence of provenance.
Not scaffolded: the comparison and the explanation.
Removal plan: next time, read aloud only the harder of the two sources.

Group C — objective: identical.
Greater depth: a third account, published forty years later, and one question: “Does distance in time make an account more reliable or less?” Same objective, more demanding application.

Notice what did not happen. Nobody got an easier source, a shorter paragraph or a different question. Group B read the same text as everyone else, because in this lesson the archaic language is the evidence.

The prompt for EAL learners

You are adapting a lesson plan for EAL learners. Language
proficiency is not an indicator of cognitive ability, and the
adaptation must not reduce demand.

Learning objective: {{objective}}
Assessed element: {{what must not be scaffolded}}
Core task: {{paste}}
Pupils: {{EAL stage per pupil or group, first languages,
  whether literate in first language, time in UK education}}
Subject vocabulary: {{words}}
Everyday vocabulary the task assumes: {{the words nobody
  thinks to teach}}

For each stage represented, output:
- The same objective.
- LANGUAGE DEMAND AUDIT: which parts of the task are
  linguistically hard rather than cognitively hard.
- VISUAL OR CONCRETE SUPPORT: what makes the concept
  accessible without words.
- TALK BEFORE WRITING: the spoken task that comes first, and
  who they do it with.
- SUBSTITUTION TABLE or sentence-level support, only if
  writing is not the assessed element.
- FIRST-LANGUAGE USE: where using their first language would
  help the thinking, and where it would not.
- OUTPUT ROUTE: how they can demonstrate the objective
  without the language barrier standing in the way.

Rules:
- Never simplify the concept. Simplify the language around it.
- Distinguish clearly between subject vocabulary that must be
  taught and idiom that should be removed.
- Do not remove subject-specific terminology. Pupils need it.
- Flag any part of the task that tests English comprehension
  when it intends to test subject knowledge.

The extension prompt

You are writing extension for pupils who have met the
objective, without moving them onto the next objective.

Objective: {{objective}}
What they have already done: {{the core task}}
Where this sits in the sequence: {{what comes next, which they
  must NOT be pushed into}}
Common misconception at this level: {{misconception}}

Produce three extension routes, all on the same objective:
1. GREATER DEPTH — a harder application of the same idea.
2. TRANSFER — the same reasoning in an unfamiliar context.
3. CRITIQUE — evaluating a flawed example, or arguing the
   opposite position.

For each: the task, what it reveals about their understanding,
and how you would know it was too easy.

Rules:
- No route may teach content from the next objective.
- No route may be "more of the same" — extra questions are
  not extension.
- No route may require resources a teacher would not already
  have.
- One route must be completable in ten minutes, for pupils who
  finish early rather than pupils who are working ahead.

The review prompt: did the objective survive?

Below is a learning objective, the assessed element, and three
differentiated versions of a lesson plan.

Audit them and output:

1. OBJECTIVE DRIFT — for each version, whether the pupil is
   working toward the identical objective. Quote any wording
   that has shifted.
2. DEMAND REDUCTION — anywhere cognitive demand has been
   lowered rather than support added. This is the failure that
   matters most; be strict.
3. SCAFFOLDED ASSESSMENT — anywhere the scaffold does part of
   the assessed thing for the pupil.
4. SUBSTITUTE ACTIVITY — anything that keeps pupils busy
   without addressing the objective.
5. NO REMOVAL PLAN — any scaffold with no stated route off it.
6. TIME MISMATCH — any group whose task clearly takes longer
   or shorter than the others.
7. IMPLICIT CEILING — any wording that assumes a group cannot
   reach the full objective.

Do not improve the plans. Only audit.

OBJECTIVE: {{paste}}
ASSESSED: {{paste}}
VERSIONS: {{paste}}

Point three is subtle and worth the run. A writing frame that supplies the causal connectives is excellent support when the objective is historical reasoning and a fatal flaw when the objective is writing causal explanation. The same scaffold, opposite verdicts, and only the objective tells you which.

Differentiating a whole unit of work

You are differentiating {{n}} lessons in a unit, for the same
class, as a sequence rather than as separate lessons.

Unit objectives in sequence: {{list}}
Assessed elements per lesson: {{list}}
Lesson plans: {{paste or summarise each}}
The class, by barrier: {{groups and barriers}}
Statutory provisions: {{EHCP requirements}}

For each lesson, produce the three-way differentiation.

Rules for the sequence:
- Scaffolds must reduce across the unit. If a group has the
  same level of support in lesson 6 as in lesson 1, flag it:
  NO FADE: {{group}} — {{scaffold}}.
- Pre-taught vocabulary should not need re-teaching every
  lesson. Track what has been taught and flag repeats.
- Statutory provisions appear in every lesson they apply to.
- Output a fade map: group down the side, lesson across the
  top, level of scaffold in each cell, so the reduction is
  visible at a glance.
- Flag any lesson where a group's objective could not be kept
  identical, with the reason.

The fade map is the output worth keeping. It makes a term’s worth of differentiation visible as a trajectory, which is the only way to see whether support is being withdrawn or has quietly become permanent.

Common mistakes when differentiating a lesson plan

Describing groups by attainment band instead of by barrier. “Middle group” produces generic support. “Cannot hold three steps in working memory” produces a comparison grid. The barrier field is where all the quality comes from.

Leaving the assessed-element field blank. Without it the model cannot tell a legitimate scaffold from one that does the task, and it will produce something that looks supportive and invalidates your assessment.

Accepting a differentiated task that takes a different length of time. Two groups finishing fifteen minutes apart is a behaviour problem you have designed into the lesson.

Never removing a scaffold. The removal plan feels like the least important field and it is the one that determines whether differentiation closes a gap or maintains it. The EEF guidance on scaffolding is consistent on this: the point is the withdrawal.

What to check before teaching a differentiated lesson plan

Read the three objectives side by side. If they are not word for word identical, you have three lessons rather than one differentiated lesson, and one group is learning less.

Check the assessed element in each version and confirm the pupil is still doing it. Then check the timing, so that all three groups finish within a few minutes of each other.

Confirm every statutory provision from an EHCP appears in the version that pupil will use. A model that was not told about a provision will produce a plan that omits it, fluently.

Finally, look at the removal plan and put the next reduction in your planning for the following lesson. Scaffolds that are not scheduled to come off do not come off.