Yes — teachers in England can use generative AI to help with marking and feedback. The Department for Education says teachers can use AI for "marking work, giving feedback", provided they "use their professional judgement and check that anything AI generates is accurate and appropriate". There are clear limits: no summative judgements by AI, thorough checking of outputs, and real care with pupil data. Here is what the documents actually say, quoted and linked.
Which documents set out the DfE's position?
Three official sources matter:
- The policy paper — Generative artificial intelligence (AI) in education, first published in March 2023 and last updated on 12 August 2025. This is the DfE's formal position statement.
- The support materials — Using AI in education settings: support materials, published on 10 June 2025 and updated on 19 May 2026. Four training modules plus support for school and college leaders. Module 4, on use cases, is where marking and feedback get the most direct treatment.
- The Education Hub explainer — AI in schools and colleges: what you need to know (10 June 2025), a plain-English summary of the position.
Does the DfE say teachers can use AI for marking and feedback?
Yes, explicitly. The Education Hub explainer answers its own question — "Are teachers allowed to use AI?" — like this:
"Yes, teachers can use AI to help with things like planning lessons, creating resources, marking work, giving feedback, and handling administrative tasks. But they need to use their professional judgement and check that anything AI generates is accurate and appropriate — the final responsibility always rests with them and their school or college."
The policy paper says "research demonstrates that generative AI could also be used for tasks such as feedback and tailored support in schools", and lists "tailored feedback and revision activities" among the uses the sector has identified. It also notes the DfE sees "more immediate benefits and fewer risks from teacher-facing use of generative AI" — tools used by the teacher, rather than put in front of pupils.
The Module 4 training materials get specific about feedback:
"For feedback, AI might help teachers draft comments on common misconceptions or suggest next steps for improvement."
The DfE is also funding new AI tools for education that "stretch the capabilities of what AI can do — from assessing handwritten work to giving feedback on hand-drawn geography maps".
What conditions come attached?
Teacher oversight is the non-negotiable thread through every document. The policy paper puts responsibility squarely on the professional:
"Teachers, leaders and staff must use their professional judgement when using these tools. Any content produced requires critical judgement to check for appropriateness and accuracy. The quality and content of any final documents remains the responsibility of the professional who produced it and the organisation they belong to, regardless of the tools or resources used."
Module 4 applies the same principle directly to marking:
"The professional judgement about each individual student's needs, progress and appropriate feedback remains firmly with the teacher, and the process of marking helps the teacher to understand their students better."
It adds that "any approach to using AI tools for marking should balance workload reduction with ensuring accuracy and enabling educators to better understand their students' progress", and flags a practical limitation of general-purpose chatbots: LLMs "are designed to give different answers when asked the same question, and so if they are used to mark work, they may produce inconsistent feedback".
Formative assessment gets a specific caution — not a prohibition:
"Formative assessment tasks are an integral part of planning the next steps in a pupil or student's learning so using AI to reduce workload here needs careful consideration."
Where does the DfE say AI should not be used?
Module 4 lists the areas where AI should not be used, under the heading of not "outsourcing high-stakes decisions to AI":
- "AI shouldn't be used to make summative judgements or predictions about student attainment or progress."
- Be careful using AI for "references or reports that will go to external parties without thorough review and adaptation".
- "AI shouldn't be used for any task involving personal student data unless you have explicit approval and appropriate safeguards are in place."
- "Most importantly, your use of AI shouldn't replace your professional judgement about what's appropriate for your specific students."
The summary line is a good one-sentence policy: "AI can support and inform your decisions, but the decisions themselves must remain yours."
What about pupil data and children's work?
The policy paper is direct: "Personal data must be protected in accordance with data protection legislation. It is recommended that personal data is not used in generative AI tools." Where it is "strictly necessary", the school or college "must ensure that all steps are taken to protect the data" and that products comply with data protection legislation and the school's data privacy policies.
Module 4 adds two points teachers sometimes miss. First, never input personal data about students — "names, assessment data, SEND information, or any details that could identify individual students" — into AI tools "unless you have been advised it is safe to do so". Second, "students own the copyright to their own work", so putting student work into AI tools without consent "raises significant intellectual property concerns".
In practice this is the difference between pasting a child's writing into a general-purpose chatbot and using a purpose-built tool with the safeguards, contracts and data-processing terms a school can actually check. We cover how Howay handles this in Is AI marking safe?
The DfE's position at a glance
| Question | What the guidance says | Where |
|---|---|---|
| Can AI help with marking and feedback? | Yes — drafting comments, suggesting next steps | Module 4, Education Hub |
| Who is responsible for the output? | The professional who produced it, and their school | Policy paper |
| Can AI make summative judgements? | No — no summative judgements or predictions about attainment | Module 4 |
| Can pupil personal data go into AI tools? | Recommended not; if strictly necessary, with safeguards and compliance | Policy paper |
| Are general chatbots reliable markers? | LLMs "may produce inconsistent feedback" if used to mark work | Module 4 |
What does this mean for a primary teacher using a tool like Howay?
The DfE's model — AI drafts, the teacher decides — is the model Howay is built around. Howay returns curriculum-aligned feedback on handwritten writing, surfaces the criteria a piece of work meets rather than putting a numeric grade on a child's work, and keeps every output editable by the teacher before it reaches a child or parent. The summative judgement about a child's attainment stays where the guidance says it must: with the teacher. You can see the full workflow in How it works.
Frequently asked questions
Is AI-assisted marking allowed in England's schools?
Yes. The DfE says teachers can use AI for "marking work, giving feedback", provided they check that outputs are accurate and appropriate — the final responsibility rests with the teacher and their school.
Can AI decide a pupil's level or predict their results?
No. The DfE's training materials state that "AI shouldn't be used to make summative judgements or predictions about student attainment or progress". AI can inform a teacher's judgement; it cannot replace it.
Do teachers have to check AI-generated feedback before using it?
Yes. The policy paper says any AI-produced content "requires critical judgement to check for appropriateness and accuracy", and responsibility for the final document sits with the professional who produced it.
Does the guidance apply to primary schools?
Yes. The policy paper covers schools and colleges in England, with a particular focus on compulsory education — primary included. The support materials are aimed at all school and college staff.