Personalized Feedback for Every Student
On this page
Feedback: A Major Pedagogical Lever
Educational research, starting with John Hattie and Helen Timperley’s 2007 synthesis, ranks feedback among the most powerful drivers of learning—provided it focuses on the task and indicates how to improve. The problem is arithmetic. With thirty students per class and multiple classes to manage, writing individualized comments on every paper is impossible. You end up writing “good job” or “needs review” at the bottom of the page—two phrases that teach no one anything. ChatGPT does not change your judgment of the work; it offloads the wording, which is the most time-consuming part.
What Makes Feedback Useful
Before handing anything over to a model, you need to know what you expect from it. Useful feedback is specific: it points out what succeeded or needs improvement, whereas “lacks rigor” leaves the student with no way forward. It is actionable; asking an eighth grader to start each paragraph with a logical connector is an instruction they can follow on the next assignment, unlike “improve your argumentation.”
It starts with the positive—what works before what is missing—not out of politeness, but because a student who feels recognized reads the rest of the comment. It is adapted to the level: the tone used with a struggling student who needs confidence building is not the same as that used with an advanced student whom you can push for higher standards. Finally, it is fast, because the sooner feedback arrives, the better the student remembers what they intended to do.
Describe the Paper, Then Let the Model Write
The model does not see the paper: you describe what you read in it. Give it a summary of strengths and errors, or a short excerpt without names or personal details, the student’s general level (struggling, average, advanced), the skills assessed, and the desired tone—encouraging, directive, or collaborative. These four pieces of information are enough.
“A 7th-grade student submitted an argumentative essay. Strengths: clear three-part structure, relevant personal examples. Weaknesses: no logical connectors, conclusion too short, two subject-verb agreement errors. Level: average. Write 6–8 lines of encouraging and constructive feedback, with one specific suggestion for each weakness.”
Notice that each weakness is paired with a request for a suggestion: this prevents the model from merely stating observations. Also notice what the prompt does not contain: no name, no full copy, no private information.
Process a Stack by Error Families
In a stack of papers, the same errors recur across ten students. Instead of rewriting the same explanation ten times:
- Identify the three or four most frequent errors in the stack.
- Ask ChatGPT to draft a template feedback for each of these errors.
- Personalize each feedback yourself, outside the tool, with the first name and an element specific to the paper.
The last step is not cosmetic: it separates the comment the student recognizes as their own from the text they might guess is generic.
Prepare the Conference, Don’t Improvise It
A well-prepared individual conference is worth three written comments. ChatGPT provides you in advance with a summary of the student’s strengths and areas for growth over a period, questions designed to trigger their metacognition—getting them to say where they are stuck rather than telling them—SMART goals to co-construct with them, and remediation activities to send home with them.
“Prepare a 5-minute individual conference with a 12th-grade student who has inconsistent results in philosophy. Their grades: 14, 8, 15, 7, 13. They master analysis but lose points on essay structure. Propose 3 open-ended questions to ask them and one realistic goal for the next assignment.”
The duration stated in the prompt forces the model to prioritize and saves you from a list of twelve points impossible to cover between two classes.
Have Students Peer-Review
ChatGPT also builds the tools for peer review. Ask it for clear reading criteria calibrated to your students’ level, guided questions that prevent them from stopping at “it’s good,” templates for positive and constructive phrasing to deliver criticism without hurting, and a verification checklist to run before final submission. By evaluating a classmate’s work using these criteria, the student develops a critical eye for their own—and your grading load lightens accordingly.
From Isolated Assignments to Period Summaries
The same logic applies at the semester scale: gather accumulated records and turn them into a readable synthesis.
“Here are a student’s grades and comments, with no name or identifying details, over the last 6 months in history-geography: [list]. Write a 5-line progress summary that highlights improvements, identifies persistent points to work on, and proposes a goal for the next term.”
This summary then feeds into report cards, parent-teacher meetings, or class councils, where you arrive with a stabilized formulation rather than an impression to reconstruct on the spot.
Your Turn
Your first three feedbacks.
- Take three recent papers from the same assignment, at different levels
- For each, summarize strengths, errors, and level, without names
- Have the tool write the feedback with one suggestion per weakness
- Compare with your own comments: what the generated version adds, and what it lacks
- Add the first name and the specific element for each paper yourself
Official sources: Getting started with ChatGPT as an educator (OpenAI Help Center) · Teacher: How to use an AI system in the course of your duties? (CNIL, France’s data protection authority)
Key Takeaways
- Effective feedback is specific, actionable, positive first, and fast
- Process papers by groups of common errors to save time
- ChatGPT can prepare your individual conferences and progress summaries
- AI-assisted peer feedback develops student autonomy
- No names or full copies in the prompt: you personalize outside the tool