Review, Adapt, and Send
Video coming soon.
Walkthrough
This is the step that matters most. Nothing goes to a student until you've reviewed it. The final decision stays with the teacher.
Read each student's feedback carefully. Check tone, specificity, quoted evidence, and the way the comment connects to the student's actual writing. When a strength comment points to something real in the draft and explains why it works, that's the kind of feedback a student can learn from.
Three editing scenarios you'll encounter on a typical class set:
-
Substance is right, tone needs adjustment. If a growth comment is correct but too blunt, soften the phrasing without weakening the feedback. Acknowledge effort where it's real. Let the student hear that the work is seen, not just the gap.
-
Next steps need sharpening. If a comment says "include more evidence," that is still too broad. Point to the paragraph, name the kind of evidence that's missing, make the revision clear enough to act on.
-
Feedback is fine as-is. If the tone is right, the specifics are there, and the grades match the standard you're applying, leave it alone. Review does not mean changing everything.
Verify each grade. A suggested score may be close, but it still needs review. Compare the criterion, the rubric descriptor, and the actual student work. If the evidence in the writing points to a different level, change it. The score generated here is a starting point, not a verdict.
Add what only you can know. A note about a student's improvement over the term. An acknowledgment of a real risk they took. A connection to a class conversation. The system can't see those things; you can. That's where feedback becomes fully teacher-shaped, not just teacher-approved.
Post when ready. Across the class, a pattern usually emerges: some feedback needs revision, some needs a personal addition, some is already strong enough to leave alone. That's normal. The point is thoughtful review, not uniform editing.
What to read or watch next
For a structured approach to reviewing AI-generated feedback once you've done it a few times, see Evaluating AI-Generated Feedback: A Five-Part Protocol — the deeper version of what you just practised.