Draft the exemplars for your task before you mark

Draft the exemplars for your task before you mark

Writing the exemplars before you mark is the single highest-leverage moderation move most teachers know about and skip, because doing it properly used to cost an hour per band per task. The moment you have a genuine A, C and E in front of you (each in the register of a real Year [X] student at that level, each keyed to your specific rubric) the marking pile calibrates in the first ten minutes, the moderation meeting stops arguing about tone and starts arguing about evidence, and the feedback comments you write on real student pieces get sharper because you have already named the moves. AI is well suited to this. It is also very good at getting it wrong in a specific way, which is worth naming up front: the default output is a first-year-university A and a two-paragraph C with the good bits removed. This prompt is written to stop that.

The prompt

You are a senior [SUBJECT] teacher and moderator in an Australian [PRIMARY / SECONDARY / TERTIARY] school. You have written, marked and moderated enough tasks of this type to know that a good exemplar is not the "perfect answer". It is a piece written in the register of a real student at a specific band, making the specific moves the rubric rewards at that band. You also know the failure mode: the A written like a first-year university essay, and the C written as the A with paragraphs deleted. You avoid both.

I am about to hand out an assessment task for [YEAR LEVEL] [SUBJECT AND UNIT]. Before I mark twenty-seven of them, I want three exemplars drafted, each aligned to a specific band on my rubric, each in the register of a real Year [X] student.

CONTEXT
- Year level: [YEAR LEVEL]
- Subject and unit: [e.g. Year 9 English, unit on argumentative writing on a set text]
- Curriculum frame: ACARA v9.0 [PASTE the specific content descriptor], or the state study design line
- The task, exactly as students see it: """[PASTE THE FULL TASK BRIEF, INCLUDING STIMULUS AND CONDITIONS]"""
- The rubric or marking guide, exactly as I will use it: """[PASTE THE RUBRIC]"""
- The label for the top band in this school: [A / VHA / Band 6 / Excellent]
- The label for the middle band: [C / SA / Band 3 / Sound]
- The label for the bottom band that still shows attempted engagement (not the "no evidence" band): [E / LA / Band 2 / Limited]

PRODUCE FOUR SECTIONS, IN THIS EXACT ORDER. NO PRE-AMBLE. NO "GREAT TASK" OPENERS. START WITH SECTION 1.

SECTION 1. THE TOP-BAND EXEMPLAR
The full piece, in the register of a real strong Year [X] student working at the top of their band. Length matches what the task asks for. Vocabulary and sentence complexity match the year level, not a graduate seminar. If the task is an essay, it reads as a strong Year [X] essay. If the task is a scientific report, it reads as a strong Year [X] report.

Underneath, one short block: THE MOVES. Three to five specific moves the piece makes that put it in the top band, each keyed to the exact criterion on the rubric that rewards it. Not "shows depth of understanding". Yes "opens with a claim that names a specific text move rather than restating the question, which the rubric rewards under criterion 2b".

SECTION 2. THE MIDDLE-BAND EXEMPLAR
The full piece, in the register of a real Year [X] student working in the middle of the class. This is the harder exemplar to write and the one AI tools most often get wrong. The middle-band piece is NOT the top-band piece with paragraphs cut. It is a different piece with a different shape. Common shapes: an idea that is genuinely there but is carried by shorter sentences, evidence that is present but not selected for the argument, a step in the reasoning that is missing rather than wrong, a structural move that hides the thinking rather than showing it.

Underneath, one short block: THE MOVES. Three to five specific moves that keep this piece in the middle band, each keyed to the rubric. Include at least one strength (something the middle-band piece is doing that a lower-band piece is not).

SECTION 3. THE LOWER-BAND EXEMPLAR
The full piece, in the register of a Year [X] student who submitted the task, attempted it in good faith, and missed the central thing it was asking for. Not blank. Not troll. A piece that shows engagement, misses the point, and has enough on the page that I have to mark it rather than return it.

Underneath, one short block: THE MOVES. Two or three specific things the piece is doing that put it here, plus one thing the student is doing right that a re-teach could build on. Name the re-teach in one line.

SECTION 4. THE COMPARISON TABLE
A short table, one row per criterion on the rubric, three columns (top band, middle band, lower band). Two or three words per cell, not full sentences. Example row for a criterion called "use of evidence": TOP: selected, integrated, cited. MIDDLE: present, listed, not selected. LOWER: paraphrased, one source, no citation.

This is the block that pays for the whole prompt at moderation, because it is what a colleague and I actually look at when we are calibrating on a shared task.

RULES
- Australian English throughout. Year 10, not 10th grade. Term 3, not Q3. Australian curriculum vocabulary.
- No em-dashes and no en-dashes anywhere. Commas and full stops only.
- Do NOT write the top band in a register above the year level. If the piece reads like a university essay, rewrite it in the voice of a strong Year [X] student.
- Do NOT write the middle band as the top band with paragraphs deleted. It is a different piece.
- Do NOT include a "here is what I have produced" opener. Start with Section 1's exemplar.
- Do NOT finish with "let me know if you would like me to write a full lesson". I will use these to mark.

What it does well

The four-section shape forces the model past its default (one A-grade sample with a friendly summary) into the three exemplars a moderator actually needs. The move that pays best is Section 2's insistence that the middle band is a different piece with a different shape, not the top with paragraphs cut, which is where most auto-generated exemplars fall over and where most moderation disputes originate. Section 4's comparison table is the artefact you will still be using in Week 8, because it is what you and the teacher next door look at when your marks on the same piece are two bands apart.

Where it falls over

The exemplars are only as good as the rubric you paste in. If the rubric is generic ("shows understanding", "uses evidence"), the exemplars will be generic too. If the criteria are named at the specific level of a student move ("selects and integrates at least two pieces of textual evidence that directly support the claim"), the exemplars sharpen dramatically. Second failure: for practical or performance tasks (VET workshop, PE skill, Drama monologue, Art folio), the top-band exemplar cannot be a written artefact and the prompt will produce a description of an exemplar rather than the exemplar itself. Read those sections as "what the teacher would look for" rather than "the piece". Third failure: for locked external assessments (VCE external exam, HSC major work) the exemplars need to sit next to the released VCAA or NESA sample responses, not replace them.

Try it like this

• For a Year 9 English argumentative task: use a real prompt from this unit, paste the analytical rubric, and check that the middle-band piece has a claim in sentence one that is genuinely a claim, not a topic. That is usually the split point in this year level.
• For a Year 10 Maths problem-solving task: paste the task and the working-and-reasoning rubric. The middle-band piece should show correct arithmetic with the reasoning implicit rather than named, which is the common Year 10 shape.
• For a VCE Unit 3 SAC: run the prompt with the study design outcome in the curriculum frame and check that the top-band exemplar is in the register of a strong Year 12 student writing under SAC conditions, not a first-year university essay. If it drifts, ask for a rewrite at Year 12 register with the specific study design vocabulary.
• If you only have twenty minutes tonight: run the prompt for Sections 1 and 4 only. The top-band exemplar and the comparison table are the block that changes tomorrow's marking. Section 2 and Section 3 can wait for the drive in.

The pile is still yours to mark. It just gets marked against three pieces you have already thought about, instead of the ghost of an A you rebuild in your head from the first six scripts you happen to open.