Build an Analytic Grading Rubric with Clear Level Descriptors for Any Assignment

You get a weighted analytic rubric with observable, non-overlapping descriptors per level, student 'I can' statements, and scoring notes for borderline work.

Prompt

Prompt
You are an experienced classroom teacher and assessment designer who writes rubrics that colleagues use to score work consistently.

Your job: build an analytic grading rubric for my assignment. Before writing anything, ask me these questions in one message, numbered, and wait for my answers:
1. What is the assignment, and what do students hand in?
2. What grade level or course is it for?
3. Which 3-5 skills or criteria matter most, and do any carry more weight?
4. How many performance levels do you want (default 4), and what is the point total?

Rules:
- Every descriptor names something a grader can see or count in the work (e.g. 'uses 3 or more pieces of relevant evidence'). Never rely on vague words like 'good', 'some' or 'excellent' alone.
- Levels do not overlap: each level differs from the one below in one clear, stated way, so a piece of work fits only one cell.
- Each criterion measures one skill only; keep grammar out of the argument row.
- Make the top level achievable at this grade level, not perfection.
- Write one short student-friendly 'I can' statement per criterion.

Output format:
1. Rubric table: Criterion | Points | one column per level, highest first.
2. 'I can' statements, one per criterion.
3. Scoring notes: how points convert to a grade, plus 2 rules for borderline work.

End with one line naming the criterion where two graders are still most likely to disagree, and why.

Example output

At a glancePoints per criterion in the 20-point essay rubricGrade 8 argumentative essay on banning phones, 4 levels
  • Claim: 4 pts
  • Evidence: 6 pts
  • Reasoning: 4 pts
  • Counterclaim: 3 pts
  • Conventions: 3 pts

Example for: a Grade 8 argumentative essay on whether schools should ban phones, 4 levels, 20 points.

1. Rubric

CriterionPoints4 – Exceeds3 – Meets2 – Approaching1 – Beginning
Claim4Claim stated in the intro, takes one clear side, and previews 2+ reasonsClaim stated in the intro and takes one clear sideClaim present but appears after the intro or takes both sidesNo identifiable claim
Evidence63+ relevant pieces of evidence, each with a named source2 relevant pieces of evidence, each with a named source1–2 pieces of evidence, at least one without a sourceNo evidence, or evidence unrelated to the claim
Reasoning4Every piece of evidence followed by 2+ sentences linking it to the claimEvery piece of evidence followed by 1 linking sentenceAt least one piece of evidence has no linking sentenceEvidence listed with no explanation
Counterclaim3Names an opposing view and rebuts it with evidenceNames an opposing view and rebuts it without evidenceNames an opposing view with no rebuttalNo opposing view mentioned
Conventions30–2 errors in spelling, punctuation or grammar3–5 errors; meaning always clear6–10 errors, or one place where meaning is unclear11+ errors, or meaning unclear in several places

2. 'I Can' Statements

  • Claim: I can state my position in my introduction.
  • Evidence: I can support my claim with facts and name where they came from.
  • Reasoning: I can explain how each fact proves my point.
  • Counterclaim: I can describe the other side and show why my view is stronger.
  • Conventions: I can proofread so my writing is easy to read.

3. Scoring Notes

Points per level: Evidence scores 6/4.5/3/1.5; Claim and Reasoning score 4/3/2/1; Counterclaim and Conventions score 3/2.25/1.5/0.75. Round the total to the nearest whole point.

TotalGrade
18–20A
15–17B
12–14C
9–11D
Below 9Not yet – revise

Borderline rules:

  1. If work meets every descriptor in a level except one, score the lower level and note the missing item in feedback.
  2. A source counts as 'named' if a reader could find it (title, author or organization); 'the internet' does not count.

Most likely disagreement: Reasoning, because graders may differ on whether a sentence truly links evidence to the claim or just restates the evidence; score two sample essays together before grading the full set.

Written by AI from this prompt. Your results will vary.

Tips and common edits

Why this works

  • Requiring countable, observable descriptors removes vague words like 'good', so two teachers land on the same score.
  • The non-overlap rule forces each level to change one stated thing, so work fits only one cell.
  • Asking for borderline rules and the likeliest disagreement point prepares teachers for moderation before grading starts.

Use when you need a rubric for an essay, project or lab report that colleagues can score the same way. Have the task sheet and grade level ready. Works in ChatGPT, Claude or Gemini.

Common edits

  • Ask for a single-point rubric version for quick feedback on drafts.
  • Request descriptors rewritten in student-friendly language for a self-assessment checklist.

Prompt details

By
Prompts-GPT editors
Published

Did this prompt work?

Keep exploring

More in Education