Interactive learning material · M4-T2-MLC2

AI-Assisted Feedback Writing for Learners

From the teacher’s observations to the learner’s next step

Author: Zsolt LengyeliTStudy Hungary Training and Research Centre
Reviewer: Mária HartyányiiTStudy Hungary Training and Research Centre

Feedback connects the learner’s current performance and next step to the learning goal. — The teacher observes, the AI drafts, the learner moves forward.

Learning objective

Formative feedback is one of the most powerful tools teachers have – and one of the most time-consuming. This material shows how AI can help teachers draft, refine and personalise written feedback on learners’ work, based on criteria the teacher has defined.

By the end of this material, you will be able to

Describe the features of effective feedback.
Write prompts that turn your own observations into clear, constructive and motivating feedback.
Adjust tone, length and language to the learner.
Review drafts for accuracy, fairness and data protection, while keeping assessment decisions in your own hands.

Introduction

Good feedback takes time

Feedback helps learning when it is specific, related to the learning goal and shows the next step. Vague praise or criticism of the learner’s personality does not provide enough guidance for improvement (Hattie and Timperley, 2007).

In a garden-building project spanning an academic year, the mentor teacher gave feedback at the end of every stage. Writing reflections for ten learners throughout the year, after each phase of work, represented an enormous workload.

AI can reduce this workload: the teacher observes and judges; the AI helps to put that judgement into clear, well-structured and encouraging words.

Learners working together on a school garden-building project. — Garden-building project. Illustration: ChatGPT.

What does this material build on?

It focuses on feedback communication (DigCompEdu 4.3: Feedback and planning) and builds on the criteria and rubrics in M4-T2-MLC1. Automatic grading and the design of adaptive learning paths are separate topics.

Effective feedback · Three questions

Where am I going, how am I going, where to next?

According to Hattie and Timperley (2007), effective feedback helps the learner answer three questions.

The learning goal, the current state of the work and the direction of improvement along a learning path. — Feed up, feed back, feed forward: the goal, the current position and the next step.

Where am I going?

What is the goal, and what are the assessment criteria?

How am I going?

Where does the work stand in relation to the criteria?

Where to next?

What should the learner do to improve?

Effective feedback · What does it address?

Talk about the work and the learning process

Feedback can address the task, the process, self-regulation and the person. An example from baking illustrates the difference.

On the task

“The centre of your pastry has sunk by about 1 cm and hasn’t cooked through completely.”

On the process

“Check the dough temperature before proving.”

On self-regulation

“Compare your work with the reference sample before handing it in.”

On the person

“You are talented.”

Effective feedback · Features

What makes feedback useful?

Criterion-based

It refers to assessment criteria the learner already knows, rather than to a general impression.

Specific

It names exactly what is good and what needs to change.

Focused

It does not list every mistake: it highlights the features that matter for improvement.

Actionable

It suggests a concrete next step the learner can take.

Timely

It reaches the learner before the next attempt, while they can still use it.

Respectful and motivating

It addresses the work in a tone that suits the learner’s age.

Formative assessment does not rank or judge, but shapes learning: it reveals successes and difficulties and makes correction possible.

The teacher-in-the-loop workflow

The teacher’s own observations are the input

AI helps put the teacher’s assessment into words. It turns our observations into structured text; the professional judgement remains ours.

The workflow ends with the learner’s next attempt: the feedback needs to be acted on.

Observation, review against the goal and follow-up are stages along the learner’s path. — The teacher observes and takes notes, the AI drafts, the teacher reviews, the learner acts.

Criteria: Start with the assessment criteria the learners already know.
Observation: Note your own judgement in keywords: two or three strengths and problems per learner, linked to the criteria.
Anonymisation: Remove names and personal details. Use a neutral label in the workflow, such as L07; do not share the information that links it back to the learner.
Drafting: Ask the AI for feedback in a fixed structure, based only on your notes.
Review: Check, correct and personalise every comment. Add what only you know about the learner to the final version you prepare locally.
Follow-up: Give learners time to act on the feedback, for example in the next practice session.

The teacher-in-the-loop workflow · Example prompt

From notes to structured feedback

Alongside the task, criteria and your own observations, specify the structure, tone and length.

Baking a plaited loaf · First-year apprentice

“You are an experienced vocational teacher of bakers. Write formative feedback for a first-year apprentice (aged 15–16) on a practical task: baking a plaited loaf. Criteria: preparing the dough, shaping, proving and baking, hygiene and organisation of the workstation.

My notes: ingredients weighed precisely; dough well kneaded; strands of uneven thickness, loose plait; proved too long, loaf flattened; workstation clean.

Structure: one specific strength; at most two points for improvement linked to the criteria; one concrete next step for the next practice. Friendly but professional tone, addressing the learner directly, maximum 90 words. Do not add anything that is not in my notes, and do not give a grade.”

Fixed structure

One specific strength.
At most two points for improvement linked to the criteria.
One concrete next step.

Clear limits

A friendly, professional tone, addressing the learner directly.
Maximum 90 words.
Based only on the notes, without a grade.

The teacher-in-the-loop workflow · Other trades

The same structure works for other tasks

The next step should relate to the professional task and be something the learner can carry out in the next session.

Electrical installation

“Label each conductor before connecting it, then check the circuit against the diagram.”

Welding

“Practise on scrap plate at a steady travel speed and compare your bead with the reference sample.”

Customer service

“Summarise the customer’s complaint in one sentence before offering a solution.”

Try it with a task of your own!

Choose a piece of learner work. Note one observed strength, one point for improvement and one concrete next step, then use these to request a short feedback draft.

Tone, length and personalisation

The same message, in a form that fits the learner

AI can adjust tone and length on request: “make it more encouraging”, “shorten it to three sentences”, “use plain language for a learner who reads slowly”, “make the next step more concrete”.

The teacher chooses the version that helps the learner and accurately conveys the original professional observation.

The same observations produce feedback in different tones and lengths; the teacher chooses. — Tone and length may change; the professional content must remain accurate.

Age and level

Simpler sentences and everyday words for younger learners; professional terminology for advanced apprentices, explained where needed.

Tone

More encouraging for a learner who has lost confidence; more direct for one who keeps repeating the same mistake. Always respectful.

Length

A short comment that the learner actually reads can be worth more than a long one they skip.

No clichés

Replace empty phrases (“Great job!”, “Keep it up!”) with concrete observations.

Tone and personalisation · Comparison

Feedback before and after

Before

“Good work, but the plaiting could be better. Pay more attention next time.”

After

“You weighed the ingredients precisely and kneaded the dough well – that is the basis of a good loaf. The plait was uneven because the strands were of different thickness, and the loaf flattened because it proved too long. Next time, weigh the strands so that they are equal, and check the proving with the finger test, not only by the clock.”

A specific strength

Naming the precise weighing and well-kneaded dough.

An explanation of the problem

Connecting the uneven plait and flattened loaf to the observed working process.

An actionable step

Weighing the strands and checking the proving with the finger test in the next practice session.

Teacher review

Review every comment!

Before feedback reaches a learner, the teacher checks every sentence. Expand the review questions below.

Steiss and colleagues (2024) studied feedback on around 200 essays by students in grades 6–12. The trained human raters’ feedback was of higher quality, especially in prioritising what matters most.

AI drafts feedback based on anonymised work; the teacher checks the text and makes the assessment decision. — Professional review of the draft is part of the workflow.

Source: Steiss et al. (2024).

Data protection and limits

Protect learners’ data

Before uploading

Remove names, identifying codes, photos showing faces and document metadata.

No sensitive data

Do not enter health information, special educational needs or family circumstances into general AI tools.

Use approved tools

Use tools approved by the school in line with the GDPR. Check whether the service uses entered data for model training.

Anonymisation also supports fairness

It reduces the risk of bias linked to learners’ names or backgrounds. Keep the list linking neutral work labels to names, and any other information that allows re-identification, out of the material shared with the AI.

Teacher responsibility

The grade is the teacher’s decision

AI helps with the wording of feedback. The teacher decides on grades and progression and remains responsible for every comment sent in their name.

Tell learners how AI supports feedback writing, in line with the school’s rules.

Make room for conversation alongside written feedback: a short personal talk helps you check whether the learner has understood the message.

Assessment and the AI Act

Annex III of the EU AI Act includes AI systems intended to evaluate learning outcomes among high-risk use cases. Classification depends on the system’s intended purpose and the conditions set out in the Regulation.

According to the Commission’s timeline, the rules for high-risk systems in education covered by Annex III apply from 2 December 2027.

AI Act · European Commission: implementation timeline

Ask about the next step!

Ask the learner to explain in their own words what they will do differently in the next practice session.

Takeaways for teachers

Five pointers for your next feedback

Answer three questions!
Where am I going? How am I going? Where to next?
Start with your own observations!
AI puts the teacher’s judgement into words. The professional observations and judgement remain the teacher’s.
Use a fixed structure!
One specific strength, at most two points for improvement, one concrete next step.
Adapt the form to the learner!
Use tone, length and language that support understanding. Avoid empty praise.
Review and retain responsibility!
Anonymise the work and review every comment. Keep grading decisions in human hands.

Individual practice · 40 minutes

Turn observation notes on three anonymised pieces of learner work into feedback. Save the prompt that works, and compare time and quality with your usual practice.

Classroom exercise · 45 minutes

In a customer-service role play, learners observe each other, formulate feedback, and choose a concrete goal for the next round.

Detailed instructions for both exercises are in the practical activities PDF available on the final page.

Self-check quiz

Check your knowledge!

Select one answer per question, then check your answers. Your self-check result appears only on this page.

1Which three questions does effective feedback answer?What grade did I get, who is best, who made mistakes?Where am I going, how am I going, where to next?How much time did I spend on the task, and how many words did I write?

2What should AI-assisted feedback be based on?Performance invented by the AI.A general impression of the learner’s personality.The teacher’s own observations linked to the criteria.

3Which request gives feedback a useful structure?One specific strength, at most two points for improvement, one concrete next step.List every mistake and write a long assessment.Say “well done” and give a grade.

4What should we do if the AI states something that is not in our notes?Keep it, because it is probably correct.Check and correct or remove the unsupported statement.Only change its tone.

5What needs to happen before uploading learner work?Enter the learner’s name to make the text more personal.Provide health and family information.Remove personal data, identifiers and metadata.

6Who is responsible for the final feedback and the grade?The teacher, who reviews and approves the feedback and makes the assessment decision.Only the AI service provider.The learner who receives the text.

Sources

References and further reading

Professional sources for this material

Hattie, J. and Timperley, H. (2007): The Power of Feedback. Review of Educational Research, 77(1), 81–112. Study DOI link.
Brookhart, S. M. (2017): How to Give Effective Feedback to Your Students. 2nd edition, ASCD.
Steiss, J. et al. (2024): Comparing the quality of human and ChatGPT feedback of students’ writing. Learning and Instruction, 91, 101894. Study DOI link.
Redecker, C. (2017): European Framework for the Digital Competence of Educators: DigCompEdu. Publications Office of the European Union. 4.3: Feedback and planning.
Regulation (EU) 2024/1689 of the European Parliament and of the Council, Annex III, point 3; Regulation (EU) 2016/679 (GDPR).
iTStudy Hungary: Innovative assessment methods, course module.
Lengyel, Zs. (2026): Concrete applications for content generation. VITA training presentation.

DigCompEdu

The digital competence framework for educators.

Framework website

AI Act

The official text of Annex III and the Commission’s current timeline.

European Commission: AI Act

VITA · Credits

Credits and copyright

Author: Zsolt LengyeliTStudy Hungary Training and Research Centre
Reviewer: Mária HartyányiiTStudy Hungary Training and Research Centre
Project ID: 2025-1-LT01-KA220-VET-000362456
Learning material: M4-T2-MLC2

Attachments

Download learning materialM4-T2-MLC2-TEXT-EN.pdf Download practical activitiesM4-T2-MLC2-EXC-EN.pdf

This work may be reused and adapted under the terms of the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).

Funded by the European Union. Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the National Agency.