Illustrative sample report
This is an illustrative example based on a fictional school. It is not a commissioned Review, and real reports are more comprehensive and specific to each school.
Should teachers use AI to help produce formative feedback?
The situation
A secondary school is considering whether to allow teachers to use AI to suggest formative feedback on student work. Some staff believe it could save time and help produce more useful feedback. Leaders are also aware of the risks: inaccurate or irrelevant comments, over-reliance on AI, the extra workload of checking suggestions and the handling of students' personal data.
The question
How can the school benefit from AI-assisted formative feedback without compromising its quality, reliability or teachers' professional judgement?
What current guidance says
Guidance checked October 2026. Dates shown are those of each publication or its last update.
- The Department for Education recognises that teacher-facing uses of AI may bring benefits, including tailored feedback, but teachers must check outputs for accuracy and appropriateness and remain professionally responsible for them. DfE, Generative artificial intelligence (AI) in education (last updated 12 August 2025)
- Schools should use approved tools, seek advice from their data protection officer and IT lead, and understand what personal data a tool processes. Student personal data should not be entered into unapproved tools. For this school, this is a compliance gate to clear before any trial begins. DfE, Generative artificial intelligence (AI) and data protection in schools (chapter updated 23 March 2026)
- For qualification assessments, JCQ states that AI cannot be the sole marker and that the assessor remains responsible for the final mark. This applies to qualification assessment, not automatically to routine formative feedback in lessons. JCQ, AI Use in Assessments: Your role in protecting the integrity of qualifications (revision effective 30 April 2025)
- For specified coursework qualifications, JCQ states that AI may not be the sole or primary means of marking candidate work. A human assessor must review the entire work and determine the mark, and subject and awarding body requirements prevail. Again, this governs those qualifications rather than all classroom feedback. JCQ, Instructions for conducting coursework (effective 1 September 2026)
Options considered
| Option | Assessment |
|---|---|
| Prohibit AI-assisted feedback | Gives a clear boundary but rules out any potential benefit before it has been tested. |
| Permit it broadly under a general "responsible use" statement | Simple to announce but likely to produce inconsistent practice and leaves data protection and quality risks uncontrolled. |
| A limited trial with safeguards | Lets the school test the educational and workload value before deciding whether to extend it. Recommended, but only once data protection and IT approval are in place. |
Illustrative recommendation
Only once an appropriate approved system, data protection checks and an agreed trial scope are in place, the school should permit a small pilot of AI-assisted feedback in defined, low-stakes formative tasks. Teachers must review every suggestion, correcting or rejecting it where necessary, before sharing it with students and remain responsible for the feedback students receive.
AI should not independently determine student attainment or qualification marks. The pilot is about whether feedback improves for students, not about adopting a particular product.
What the school should do next
- Agree the types of task included in a small pilot and where AI-assisted feedback is unsuitable.
- Confirm approval from the data protection officer and IT lead, the approved system and lawful handling of any student personal data. Do not begin the pilot until the data protection officer and IT lead have approved both the system and the intended data handling. Anonymising work does not make an unapproved tool acceptable. Use genuinely anonymised or sample work where practicable, and never upload identifiable student work without the school's explicit authorisation and compliance with its data protection requirements.
- Explain what staff are expected to check, with a worked example of acceptable and unacceptable use.
- Run a small trial with volunteer teachers using existing tasks, without creating new student work.
- Review the quality of feedback students receive, the time it takes including checking, and how useful students find it. Then decide whether to continue, adapt or stop.
How would the school know whether it was working?
- Teachers can explain and justify the feedback they give.
- A sample of feedback is checked for accuracy, specificity and usefulness.
- There is evidence of students acting on the feedback.
- Teachers report the net time involved, including checking, rather than assuming AI saves time.
- If checking takes disproportionate time or feedback becomes worse, the approach is changed or stopped.
The value of the Review
Rather than a list of concerns, the school receives a reasoned recommendation with safeguards, staged actions and proportionate ways to judge success. This recommendation is fictional, and a real one would depend on the school's own context.
This is a simplified, fictional example to illustrate how a Review works. A commissioned School AI Approach Review provides a more comprehensive analysis, recommendations and practical next steps based on your school's circumstances and the guidance and evidence relevant to its question.
About the School AI Approach ReviewDecisions in Practice
School AI Approach Review
Leadership Report

Discuss a Review
Every Review is scoped around one question your school agrees.