DfE guidance suggests schools use AI to draft SEND support plans

Schools can use AI to draft support plans and staff guides for pupils with SEND, updated government-backed guidance suggests.

But leaders must be “extremely cautious” with pupil data and should not outsource key decisions to the technology, nor let it write statutory documents like education, health and care plans (EHCPs).

Guidance modules on AI use in schools have been updated to include new sections on “business and operations”, “inclusion and accessibility” and “administrative tasks”, as well as expanded content on safeguarding.

The guidance was first introduced in June 2025, and has so far been downloaded more than 44,000 times.

It initially set out how AI could be used for giving feedback for pupils, writing emails and letters to parents and adapting materials for SEND pupils.

It also gave advice about deepfakes, responding to online risks, safeguarding and privacy.

It was developed by the Chiltern Learning Trust (CLT) in partnership with the Chartered College of Teaching, working closely with the DfE.

Christian Turton, ed tech lead at Chiltern Learning Trust and lead developer of the guidance, said: “AI should be a tool to support learners who face the greatest barriers – not one that widens existing gaps.

“These materials ask every school and college to think carefully about who benefits and who is left behind, and to put the needs of the learner first.”

Alison Peacock, chief executive of the Chartered College of Teaching, added that the guidance “will support all staff working in early years settings, schools and colleges to navigate the fast-moving landscape of AI in education and the risks and opportunities it brings”.

The renewed guidance was based on a survey of more than 500 teachers, and includes more than 20 new practitioner interviews, new research articles and case studies.

Here are some of the main updates:

1. Drafting support plans for SEND pupils

Teachers, teaching assistants and SENCOs can use AI to draft initial versions of support plans, create draft staff guides for supporting specific needs, or generate template letters and communications, the guidance states.

The update builds on the pre-existing guidance, which set out ways of using AI to personalise learning for SEND and neurodivergent pupils and adapt to their needs.

But the new guidance stresses that staff should be “extremely cautious” about inputting information about a child’s SEN, disabilities and additional support needs, and warns that even seemingly anonymous information can risk identifying a pupil if it is matched with other information.

It adds that some decisions should not be outsourced to AI, such as deciding whether a student has a special educational need.

Other tasks that should not be left to AI include writing statutory documents like EHCP contributions without a significant professional review, any application that has sensitive personal information without appropriate safeguards in place or for providing social, emotional and mental health support.

It also notes that existing tools, such as voice typing, screen readers, text-to-speech and colour overlays should not necessarily be overlooked in favour of new AI-tools.

Staff are also being advised to support vulnerable children in understanding the distinction between AI and humans, due to the risk that they could become emotionally attached to an AI chatbot.

2. Translating learning materials

The new content goes into depth about how AI can be used to translate material for students who are speakers of English as an additional language (EAL).

It advises the level of support should be tailored to pupils’ English proficiency and their overall literacy levels in their native language.

While some pupils will need the whole text to be translated, others may only need a crib sheet for more complex words and key vocabulary.

But schools should have translations checked by a fluent speaker where possible or translate the content back into English using several tools.

3. Risk assessments

The previous guidance touched on using AI for writing policy, but the updated guidance goes into more detail.

For example, the guidance states AI could be used to compare current school policies against updated statutory guidance, make recommendations for improvements and identify gaps in safeguarding and operational procedures.

Other tasks that can be delegated include creating risk assessments for school trips, generating itineraries, and supporting estates coordination and development planning.

But the guidance stresses that final decisions and responsibility for risk assessments and safeguarding policy must remain with the school.

AI can also be used to review contract documentation and procurement processes, create standard operating procedures from voice recordings or notes, support crisis response planning, and creating staff briefing documents for inspection readiness.

The guidance also sets out how AI can be used to create business cases, carry out pay reconciliation between budgets software and payroll and analyse staff costs, budgets and other costs.

It gives an example of a prompt asking AI to help draft a budget for next year based on the last three years’ budgets and current staff costs.

AI could also be used for data analysis – including exploring attendance and attainment patterns, creating data visualisations, and analysing anonymous survey responses from pupils, parents or staff.

4. Summarising meeting notes and staff updates

The updated guidance states AI tools can be used to summarise meeting minutes and extract “key decisions and actions”.

But schools should ensure all participants in the meeting know that they are being recorded and that what they say will be processed by AI. Any AI-generated summaries should be checked for accuracy before sharing them, the guidance adds.

The tool being used to record the meeting must also meet the school’s data protection requirements, and it should not learn from the data output.

AI can also be used for generating “consistent staff newsletters and procedural updates” and translating messages from multilingual families, the guidance states.

But while AI can help with structure and check existing messages are accessible, clear and consistent, it should not be used for messages that involve “concern, reassurance or a sensitive context”.

5. Don’t bypass pupils’ thinking

The section of the guidance on safeguarding has been expanded to include the risks of cognitive offloading – which includes pupils deferring their thinking to AI at the expense of their own educational and intellectual development.

It also acknowledges that it can be difficult to detect whether AI has been used, and that AI detection tools can be unreliable and produce “false positives” that disproportionately affect certain pupils, including EAL pupils.

It notes that the focus “must remain on ensuring students and pupils develop their own knowledge and skills rather than becoming over-reliant on AI-generated content”.

The guidance also invites teachers to reflect on how to design assessment tasks to “encourage genuine learning” and “reduce the temptation or effectiveness” of using AI.

“By fostering a culture of integrity and setting clear expectations, schools and colleges can ensure AI does not compromise authenticity or bypass the important thinking that is part of the learning process.”