The Assignment Feedback tool grades every submission in a batch against the same rubric and drafts the feedback for you to review. Each learner gets a scorecard (one row per criterion with the evidence the judgement rests on), a short summary, strengths, improvements, and next steps. You edit what needs your judgement and send it on.
What it accepts
| Input | Formats |
|---|---|
| Assignment description | A sentence or two typed on the page |
| Rubric or marking scheme | Upload a PDF, DOCX, TXT, or photo; paste the text; or let the tool draft criteria from the assignment description |
| Submissions | PDF, DOCX, PPTX, TXT, or photos of handwritten work (JPG, PNG), or text pasted straight in |
Scanned pages and photos are read with text recognition first, and you can see the recognised text before anything is graded. A batch holds a limited number of submissions; the counter on the page shows how many you have added.
Credits
Grading costs 5 credits per submission. The Generate feedback button shows the total for the batch before you start, for example "Generate feedback for 24 submissions · 120 credits." Reviewing, editing, and exporting the results cost nothing.
The problem it solves
Grading thirty essays against a rubric means holding the same criteria in mind thirty times, quoting the right line from each paper, and still writing something the learner can act on. The consistency slips by paper twenty, and the written feedback is usually the first thing cut when time runs out.
The tool keeps the criteria fixed for the whole batch. You confirm them once, every submission is judged against that same structure, and the written feedback comes out already tied to the criteria rather than as general writing advice.
What you get for each learner
- Scorecard. One row per criterion with the points or level awarded, a short quote from the submission that justifies it, and a comment.
- Summary. Two to four sentences addressed to the learner.
- Strengths, improvements, next steps. Concrete and tied to the criteria, not generic praise.
- Inline comments. A handful of comments anchored to passages in the work, about content and reasoning rather than spelling.
- Class view. The average, the spread, and the criterion most learners missed, plus a class CSV with one row per learner.
Recommended workflow
- Describe the assignment. A sentence or two about what learners were asked to do makes the feedback specific to the task.
- Add the rubric. Upload a file, paste text, or draft criteria from your task description. Choose Points or Levels scoring. A marking scheme that lists expected answer points with marks works as well as a levelled rubric.
- Confirm the criteria. The tool structures the rubric into criteria and maximum points. Fix anything it misread before grading starts.
- Upload the submissions. Check the recognised text on scanned work.
- Generate feedback and review each scorecard. Adjust marks, rewrite a comment, then copy the feedback for the learner or export the class CSV.
The goal is not to send AI feedback directly to learners. It is to reach a consistent, rubric-based judgement for every learner faster, with the written explanation already drafted.
Why the draft is a starting point
The assistant writes from what is on the page. It does not know that a learner has been stuck on one concept for weeks, that a late submission was agreed in advance, or how your department words a borderline grade. Treat every scorecard as a first pass to edit rather than a grade to forward unread, and keep the final decision yours.
Who benefits most
- Teachers grading large cohorts. The consistency and time saving grow with the number of submissions.
- Teachers marking against an answer scheme. Expected points are matched and counted for each answer.
- Teaching assistants. A confirmed rubric keeps less experienced graders on the same standard.
- Corporate trainers. Written deliverables in a programme can be judged against the same criteria for every participant.