ChatGPT Rubric Review: Check Your Assignment Against the Criteria
Compare your assignment draft with its rubric. Locate evidence, distinguish missing material from partial coverage, and choose revisions you can make yourself.
Use the rubric to choose your next revision
“Improve my essay” often produces broad advice about style. A rubric review has a narrower job: connect each assessment criterion to something in your draft, identify the gap, and help you decide what to revise. You need the actual assignment instructions and your own writing for that comparison.
ChatGPT’s feedback is a second reading, not a predicted mark. This workflow asks for evidence from numbered paragraphs and keeps missing information visible. It works for an essay, a report, or another assignment with written criteria; adjust the checks to the task rather than inventing a generic grading scheme.
1. Supply the original criteria and the review scope
Before using AI feedback, check the permitted use for this assignment. Then paste the relevant instructions and the rubric as supplied, including weights and level descriptions if they exist. Give criteria IDs such as C1 and paragraphs IDs such as P1. Do not remove a criterion because you expect the AI to understand it from the title.
State whether you are providing the full draft or only an excerpt. If the bibliography is outside that excerpt, “no citation found here” does not establish that the entire assignment has no bibliography. Remove personal identifiers and material you should not upload.
- Include the required task, length, and evidence rules when provided.
- Keep the original wording of the criteria available for comparison.
- List the sources you have actually checked; do not ask the model to fill in invented references.
- Ask for feedback and questions that leave the revision work with you.
2. Practice with a fictional rubric and draft
The following criteria are invented for this demonstration, not taken from a real course. The example paragraphs are also fictional. The weights are useful for explaining revision priority; they do not authorize the AI to calculate a grade.
| Criterion | Requirement | Weight |
|---|---|---|
| C1 | State a clear position. | 30% |
| C2 | Use two pieces of evidence and explain their limits. | 30% |
| C3 | Address one counterargument and respond to it. | 20% |
| C4 | Make the sources of evidence traceable. | 20% |
- P1: “The town should add shade at busy bus stops.”
- P2: “On Tuesday at 16:00, I observed six waiting passengers; three stood beneath the only tree.”
- P3: “However, a shelter costs money.”
- No further paragraphs or source details are supplied in this practice excerpt.
3. Ask for criterion-to-passage evidence
Open the shared prompt and insert the actual instructions, rubric, and numbered draft. Ask the reviewer to quote a short, exact passage for each judgment. “Partly supported” should explain which part is present and which is missing. “Cannot assess” is appropriate when the supplied material does not let the reviewer decide.
For the fictional excerpt, C1 has an identifiable position. C2 has one observation, but not two pieces of evidence or a discussion of limits. C3 raises a cost objection without a response. C4 cannot be fully assessed without the source information. These are comments on the supplied excerpt, not a mark for a complete essay.
Open, edit and copy the full prompt
4. Check the review before acting on it
A useful response should resemble the evidence map below. If the model quotes a sentence you never wrote or claims that C3 requires two objections, compare its output with the original text and correct that mismatch before revising.
| Criterion | Evidence in the excerpt | Gap to examine |
|---|---|---|
| C1 | “The town should add shade” [P1]. | The position is stated; check its fit with the actual assignment. |
| C2 | One observation of six passengers [P2]. | Find another relevant, checked source and explain what the observation cannot show. |
| C3 | “a shelter costs money” [P3]. | The objection is present; a supported response is not found in the excerpt. |
| C4 | No source details supplied. | Cannot fully assess traceability; inspect the source notes or bibliography. |
5. Revise the evidence before polishing the wording
The sample observation describes six people at one place and one time. It does not show what all passengers prefer or whether shade caused their choices. A useful review should notice that limit rather than converting the observation into a universal claim.
For this example, you could first inspect C2 because the 30% criterion has a substantial gap, then develop the response required by C3, and check source traceability alongside both. That order is an editorial suggestion based on the supplied weights and gaps, not a calculated grade. If the real task has different priorities, follow its criteria.
Ask questions that support your own revision: “What other explanations could fit P2?” “What checked evidence would help me assess the cost objection?” “Which sentence extends beyond what my evidence can support?” Gather the needed material yourself before writing the revised passage.
6. Recheck the revised passages
After revising, keep the same paragraph IDs and ask for a limited second review of the changed sections. Require the reviewer to distinguish resolved gaps from remaining ones. Do not accept a general “everything meets the rubric” without an evidence map.
With 20 minutes available, a possible split is five minutes to read the feedback, ten to revise the most consequential gap, and five to recheck the affected criteria. Source research may require a separate session. A time limit should narrow the revision, not justify invented evidence.
- Every quoted passage exists in the supplied draft.
- Criterion wording, weights, and review scope match the original task.
- Observations stay distinct from wider claims and their limits are explained.
- New evidence and citations have been checked at their source.
- The final wording is yours and follows the assignment’s rules.
Related guides and source note
OpenAI notes that ChatGPT can make factual errors. Verify its quotations and source claims against the materials you provided. The fictional rubric and review method in this guide are Praclane’s worked example, not an official grading framework.
Photo: lilartsy / Unsplash.
