What should a plenary do in a lesson?

A plenary should assess whether the lesson's learning objective was met — not summarise what was covered. An exit ticket with two or three questions mapped directly to the objective gives the teacher actionable information about learning rather than engagement. 'Ask students what they learned' checks participation, not learning. The plenary is the lesson's quality control mechanism; it only works if it's designed to test the specific outcome the lesson was trying to produce.

Why AI Lesson Plans Fail the Learning Objective Test

Q: What is backward design in lesson planning?

Backward design, developed by Wiggins and McTighe, starts with the desired learning outcome and works backward to design instruction. Instead of planning activities and then attaching an objective, the teacher defines what students should be able to do at the end of the lesson, designs an assessment that would confirm they can do it, and then builds activities that produce that specific learning. Most AI lesson plans work forward — activities first, objective as a label — which produces plans that cover content without reliably producing learning.

Q: How does the Lesson Plan Builder skill differ from asking Claude for a lesson plan?

Asking vanilla Claude produces a generically structured lesson plan — correct sections, appropriate vocabulary, no specificity to your class, concept, or objective. The Lesson Plan Builder skill asks about your year group, learning objective, and students' prior knowledge before writing anything, then researches current evidence-based approaches to teaching that concept at that level. The result is a backward-designed plan where every activity points at a defined outcome, differentiation names specific tasks, and the plenary directly assesses the objective.

Quick answer: Most AI lesson plans list a learning objective at the top and then plan a completely separate lesson underneath it. The objective becomes a label rather than a design brief — and that gap is why the plan looks right on paper and struggles in the room.

In this guide

Lesson Plan Builder is a Claude AI skill — backward-designed plans where every activity points at the same defined outcome.

Forward Design vs Backward Design — What Actually Differs
What Makes a Learning Objective Actually Usable
The Misconception Problem Generic Plans Don't Address
What Specified Differentiation Actually Looks Like
Questions About AI Lesson Planning

Run a simple test on any AI-generated lesson plan. Read the learning objective. Then read each activity in the plan. For each one, ask: is this activity specifically designed to produce the learning the objective describes? In most AI lesson plans, the honest answer is: not really. The starter is a reasonable starter. The main activity covers the topic. The plenary asks students to reflect. All of that could have been planned without ever reading the objective — and in most cases, it was.

This is the lesson objective test, and generic AI lesson plans fail it almost universally. They produce plans where the objective is stated correctly and then treated as a heading rather than a constraint. The plan is assembled from appropriate-looking parts, but those parts aren't coordinated toward a specific destination because the planning process didn't work backward from one.

The consequence is felt in the classroom rather than on paper. The plan reads well in a review. The lesson proceeds through its stages. Students do the activities. And at the end, the teacher assesses and finds that some students understood and some didn't — but can't trace the gap to a specific planning decision, because the plan didn't make specific enough claims about how the learning would happen. That lack of traceability is a structural feature of forward-designed plans, not a failure of execution.

Forward Design vs Backward Design — What Actually Differs

Most AI lesson plans are forward-designed: topic first, then activities that cover the topic, then an objective added at the top to describe what was covered. Backward design reverses this sequence. It starts with the specific outcome — what students will be able to do, demonstrably, at the end of the lesson — and works back to design activities that produce exactly that outcome.

The difference isn't philosophical. It shows up in every component of the plan.

Plan element	Forward-designed (generic AI)	Backward-designed (objective-led)
Learning objective	"Students will understand photosynthesis." Describes topic coverage, not an observable outcome. Cannot be directly assessed.	"Students can write the photosynthesis equation and explain the role of chlorophyll." Specific, observable, assessable at lesson end.
Starter activity	General activation: "What do you know about plants?" Retrieves broad associations rather than the specific prior knowledge this lesson depends on.	Targeted activation: card sort that surfaces the specific misconception (light as raw material vs. energy source) that will silently undermine the lesson if unaddressed.
Main activity	Covers the content: worksheet on inputs and outputs. Produces content familiarity, not necessarily the specific understanding the objective names.	Produces the specific learning: annotated equation build where students predict each component and justify — directly targets the conceptual understanding the objective requires.
Differentiation	Category-level: "Provide visual aids for support; extension questions for higher ability." Describes a type of action; teacher must generate all actual content.	Activity-level: three tiered cards — scaffolded completion, independent explanation, experiment design — each mapped to a defined level of the objective.
Plenary / assessment	"Ask students to summarise what they learned." Measures engagement and recall breadth, not whether the objective was met.	Exit ticket: three questions mapped directly to the objective's components. Teacher leaves the lesson knowing exactly which students met it and which didn't.

The forward-designed plan produces a lesson. The backward-designed plan produces the specific learning it names — or at minimum, gives the teacher clear information about where the gaps are. That's the difference a well-designed objective actually makes when it drives the plan rather than labels it.

That gap is exactly what the Lesson Plan Builder skill for Claude was built to close.

What Makes a Learning Objective Actually Usable

The learning objective is the plan's most load-bearing component and the most frequently underwritten one. A vague objective doesn't just fail to guide the planning — it actively conceals whether the lesson worked. "Students will understand the causes of World War One" cannot be assessed at the end of a lesson. There is no observable student behaviour that confirms or refutes "understanding" without further specification. A teacher who writes this objective has given themselves no test to apply to their own lesson.

A usable objective names a specific, observable action students will be able to perform. Bloom's Taxonomy provides the practical vocabulary: recall, explain, compare, apply, analyse, construct, evaluate. Each describes a cognitive operation that can be demonstrated and assessed. "Students can explain two economic causes of World War One using evidence from the July Crisis" is assessable — you can write an exit ticket for it in two minutes. You know what a successful answer looks like. You know what activity produces that kind of explanation. You can design backward from it.

💡

The usability test

A learning objective is usable if you can write a two-question exit ticket for it right now. If you can't — if the objective describes a state of mind ("understand," "appreciate," "be aware of") rather than a demonstrable action — rewrite it before planning the lesson.

Generic AI lesson plans fail the usability test because they're trained on a large sample of lesson plans that collectively include a high proportion of poorly written objectives. "Students will learn about" and "Students will understand" are statistically common in the training data, so they appear frequently in the output. The Lesson Plan Builder skill pushes back on this: if the objective you provide isn't specific enough to design backward from, the skill asks you to sharpen it before proceeding. The objective is the brief. It has to be clear before the plan can be useful.

The Misconception Problem Generic Plans Don't Address

Every well-researched topic at every age group has a set of predictable misconceptions — beliefs students reliably arrive with that will interfere with the new learning if they aren't surfaced and addressed. These are documented in subject-specific pedagogical literature for most mainstream curriculum topics. They're also invisible to generic AI lesson plans, which plan activities that deliver correct information but don't account for the incorrect information already in students' heads.

The misconception that light is a raw material consumed by photosynthesis rather than an energy source that drives it affects a significant proportion of students approaching this topic at Key Stage 3. A lesson that explains the photosynthesis equation without first surfacing and addressing this misconception will find that some students leave the lesson with a revised version of the misconception rather than a corrected understanding — because new information gets assimilated to existing beliefs when the existing beliefs aren't made explicit and challenged first.

The starter activity that surfaces the right misconception does more teaching than an explanation that assumes the misconception doesn't exist.

The Lesson Plan Builder skill builds misconception-addressing into the starter by design. Before producing any plan, it researches the common misconceptions for this concept at this level — and designs the starter to make those misconceptions explicit before the lesson proceeds. This is the difference between a starter that activates prior knowledge generally ("what do you know about plants?") and one that activates it specifically and strategically, creating a cognitive conflict that the main activity then resolves.

NovaKit Skill · Educator Bundle

Lesson Plan Builder — backward-designed, objective-led, misconception-aware

Works inside Claude. No setup. Complete lesson plans where every element is designed toward a specific, assessable outcome.

See the skill $8 · instant download

What Specified Differentiation Actually Looks Like

Differentiation is the section of the generic AI lesson plan that does the least work for the most words. "Provide visual aids for lower-ability students; challenge higher-ability students with extension tasks" appears in some form in nearly every AI-generated lesson plan. It's not wrong. It's just not a plan — it's a reminder that differentiation should exist, which every teacher already knows.

Usable differentiation specifies what different learners will do differently at the activity level. Tiered activity cards are one practical structure that works across most subjects and age groups. A scaffolded tier provides the cognitive support that allows students working below the target level to access the core activity — this might be a sentence frame, a partially completed diagram, a word bank, or a simplified version of the central task. A core tier is the main activity as designed. An extension tier adds complexity that requires students to apply, analyse, or evaluate rather than just recall or explain.

The skill produces differentiation at this level of specification because it knows the specific concept and the specific year group — and can therefore name what "scaffolded," "core," and "extension" mean for this lesson, rather than describing the categories in the abstract. A teacher picking up the plan knows exactly what to prepare and when to deploy each version. No additional invention required.

Questions About AI Lesson Planning

Why do AI-generated lesson plans feel hollow in the classroom?

AI lesson plans feel hollow because they're structured correctly but not designed toward anything specific. The sections are present — starter, main activity, plenary, differentiation — but they're not coordinated around a specific learning objective for a specific group of students. The result is a schedule with educational vocabulary rather than a plan that moves learners from prior knowledge to a defined outcome.

What is backward design in lesson planning?

Backward design starts with the desired learning outcome and works backward to design instruction. Instead of planning activities and then attaching an objective, the teacher defines what students should demonstrably be able to do at the end of the lesson, designs an assessment that would confirm they can do it, then builds activities that produce that specific learning. Most AI lesson plans work forward — activities first, objective as a label — which produces plans that cover content without reliably producing the intended learning.

What makes a good learning objective for a lesson plan?

A useful learning objective names a specific, observable action students will be able to perform at the end of the lesson — not a topic they will "understand." "Students can explain the difference between mitosis and meiosis using a labelled diagram" is usable. "Students will understand cell division" is not — it cannot be directly assessed and doesn't tell the teacher what activity would produce the learning. Bloom's Taxonomy action verbs are the practical test: explain, analyse, construct, evaluate.

How should differentiation work in a lesson plan?

Effective differentiation specifies what different learners will actually do differently — not what category of support they receive. "Provide visual aids for lower-ability students" is a category, not a plan. A usable differentiation note names the specific task at each tier: scaffolded (sentence frame, word bank, partially completed diagram), core (the main activity), and extension (application, analysis, or evaluation that goes beyond the core objective). The teacher should be able to prepare the materials directly from the differentiation section without additional invention.

What should a plenary actually do?

A plenary should assess whether the lesson's learning objective was met — not summarise what was covered. An exit ticket with two or three questions mapped directly to the objective gives the teacher actionable data about learning. "Ask students what they learned" measures participation, not learning. The plenary is the lesson's quality control mechanism; it only works if it's designed to test the specific outcome the lesson was built to produce.

How does the Lesson Plan Builder skill differ from asking Claude for a lesson plan?

Asking vanilla Claude produces a generically structured plan — correct sections, appropriate vocabulary, no specificity to your class, concept, or objective. The Lesson Plan Builder skill asks about your year group, learning objective, and students' prior knowledge before writing anything, then researches current approaches to teaching that concept at that level. The result is a backward-designed plan where every activity points at a defined outcome, differentiation names specific tasks, and the plenary directly assesses the objective.

A lesson plan that doesn't commit to a specific outcome can't be evaluated against one. That's not a feature — it's the mechanism by which mediocre planning stays invisible. When the objective is specific, the lesson either produced the learning or it didn't, and the teacher knows which. That accountability is uncomfortable and also the only way to get better at the job. Generic AI lesson plans are structured to avoid it. A calibrated skill is designed to require it.

The next piece most people tackle from here is exam questions calibrated to the right level and format. If you're working across the full Educator workflow, the Educator bundle covers everything in one place.

Ready to try it?

Lesson Plan Builder for Claude

Backward-designed lesson plans with specified differentiation, misconception-addressing starters, and objective-linked exit tickets. Works with your free Claude account.

Get the skill $8 · instant download · 7-day refund

Put this to work: the Lesson Plan Builder skill for Claude turns everything above into one guided workflow you run in a normal Claude chat. Not ready to buy? Start with a free Claude skill and see how it works first.

Tags Lesson Planning Backward Design Teacher Tools Claude AI Educator Bundle