Latimer Tuition

Navigation

Latimer Tuition

Tutor professional practice

Using AI for lesson planning without risking pupil confidentiality

A practical UK guide to using AI for lesson structures, activities and explanations while keeping identifiable pupil data, SEND details and safeguarding-adjacent information out of everyday AI use.

Become a Latimer tutor

Editorial

  • Portrait of Robin GibbonsLead authorFounder & CEO·Latimer Tuition
  • Reviewed byHeather LatimerTutor Matching & Customer Service·Latimer Tuition
  • Published
  • Last reviewed

Where AI can help with lesson planning

AI is most useful when it works from curriculum and teaching information rather than pupil case material. For example, a safer request might be: “Plan a 60-minute GCSE science revision lesson on osmosis with a retrieval starter, common misconceptions, one worked example and two exit tasks for mixed-confidence learners.” That gives the tool useful teaching constraints without naming or describing a real pupil.

Recommendation

Lesson structure

Create a lesson sequence from a topic, level, time limit and objective. Check that the order, pace and tasks match the pupil’s actual needs before teaching it.

Recommendation

Retrieval and quiz questions

Generate quick checks, hinge questions and exit tasks from a topic or specification area. Review the answers and remove anything ambiguous or off-level.

Recommendation

Alternative explanations

Ask for a simpler, more visual or more step-by-step explanation of a concept without naming the pupil who needs it.

Recommendation

Differentiation ideas

Use broad, non-identifying learning needs such as lower confidence, high reading load, need for worked examples or more scaffolded practice.

Recommendation

Resource adaptation

Turn a generic task into consolidation, extension or revision practice, then check subject accuracy and suitability.

Current answer

Can tutors use AI for lesson planning without risking confidentiality?

Yes — tutors can use AI for lesson-planning support when the input is generic or genuinely anonymous: topic, level, lesson goal, time available, likely misconceptions and teaching ideas. The safer rule is simple: keep identifiable pupil data out of ordinary AI use.

The ICO explains that personal information can identify someone directly or indirectly. It also makes clear that pseudonymised information can still be personal data, so deleting a name is not enough if the remaining details still point to one child.

For lesson planning, apply the ICO’s data-minimisation wording: personal data should be “adequate, relevant and limited to what is necessary” — ICO. In most everyday planning, the necessary amount of identifiable pupil data is none.

If a task ever involves identifiable pupil data, school records, SEND or health details, or safeguarding-adjacent information, treat it as a different level of risk. It needs a clear purpose, a lawful basis before processing starts, documentation, privacy information and stronger security controls. AI output should also be treated as a draft that a tutor checks and adapts before using with a pupil.

Safe, risky and unsafe AI inputs for tutors

Use this as a practical risk check before entering lesson-planning information into an AI tool. Lower risk does not mean no risk, but generic teaching details are very different from identifiable pupil records.

A risk table comparing generic curriculum inputs, broad learner characteristics, pupil-specific records, sensitive details and direct identifiers.

Input typeRisk levelExamplesSafer alternative

Curriculum and teaching objective

Lower risk

Topic, level, curriculum aim, misconception, time limit or desired activity type.

Use these as the main planning inputs.

Broad learner characteristics

Usually manageable if non-identifying

Mixed confidence, low prior knowledge, high reading load, needs more worked examples.

Keep details general and reusable across many learners.

Pupil-specific academic history

Risky

A named report, exact marks tied to a child, school comments or a unique subject combination.

Abstract the issue into a general teaching need.

SEND, health, wellbeing or family details

High risk

Diagnosis, EHCP extracts, mental-health notes, family circumstances or safeguarding-adjacent context.

Use broad teaching supports unless a verified lawful, secure and documented process exists.

Names and direct identifiers

Unsafe for ordinary use

Name, email, school, address, phone number, full timetable or parent communications.

Do not include them in ordinary AI lesson-planning use.

Key confidentiality terms tutors need to know

These terms decide whether AI lesson planning stays in a low-risk teaching space or becomes regulated personal-data processing. The practical test is whether a real pupil can still be identified, directly or indirectly.

Personal data

Information relating to an identified or identifiable person. A pupil may be identifiable from a combination of details, even without a name.

Special category data

More sensitive personal data that needs stronger protection. The ICO lists health data in this category; tutoring details about SEND, wellbeing or safeguarding-adjacent matters can quickly become high-risk depending on context.

Anonymisation

Changing information so the person is no longer identifiable by reasonably available means. Removing a name alone may not be enough.

Pseudonymisation

Replacing direct identifiers with labels or codes. This can reduce risk, but pseudonymised information is still personal data if the person can still be identified.

Data minimisation

Using only the personal data needed for the purpose. For AI lesson planning, that usually supports generic lesson details instead of live pupil records.

Lawful basis

A valid legal basis for processing personal data. The ICO says it should be chosen before processing starts and documented.

DPIA

A data protection impact assessment. It is used to identify and reduce privacy risks in higher-risk processing, including some AI uses.

AI-specific privacy risks

The ICO highlights risks such as model inversion and membership inference. Tutors do not need to become technical specialists, but these risks explain why minimising pupil data matters.

Passkey

A modern login method recommended by the NCSC where supported. It can help reduce phishing risk and forms part of secure professional account use.

How to turn pupil detail into a generic teaching need

A safer planning pattern is to keep the educational value while removing information that points to one real child. Abstraction only works if the remaining details cannot realistically identify the pupil.

Examples showing how to replace identifiable pupil detail with general teaching requirements.

Avoid enteringUse insteadWhy it is safer

A named Year 8 pupil with dyslexia, a recent exclusion, timed-writing panic and a small-school context.

A Year 8 learner who needs low-reading-load instructions, chunked writing tasks, confidence-building retrieval and reduced time pressure.

It focuses on reusable teaching support rather than a case history that could identify one child.

A pupil report with teacher comments, exact scores and school identifiers.

A learner working below target on algebraic manipulation who needs three scaffolded examples and a short diagnostic starter.

It keeps the planning value while removing direct and indirect identifiers.

Parent messages about a pupil’s anxiety, family situation and target school.

A revision lesson for a learner who benefits from low-stakes retrieval, short tasks and clear success criteria.

It turns sensitive context into a general teaching design requirement.

What to keep out of everyday AI lesson planning

Keep these details out of ordinary AI use unless a verified lawful, secure and documented process exists for that exact situation.

  • Direct identifiers

    Names, email addresses, phone numbers, home addresses, schools, full timetables and parent contact details.

  • Reports and messages

    School reports, parent emails, teacher comments, attendance explanations, behaviour incidents and notes tied to a real child.

  • Marked work linked to a pupil

    Marked scripts, exact scores or feedback where the child can be identified directly or indirectly.

  • SEND, health and wellbeing details

    EHCP extracts, diagnoses, health information, mental-health notes, wellbeing records or family circumstances.

  • Safeguarding-adjacent information

    Any detail about risk, disclosure, welfare concerns or sensitive family context that should not be handled through a lesson-planning tool.

  • Identifying combinations

    Small clusters of details that point to one pupil even if no single field names them.

  • Stricter local rules

    Anything a school, agency, parent agreement or organisation you work with says must not be shared with an external tool.

Check every AI-generated lesson before using it

AI output is a first draft. Oak National Academy says its education AI support is “not perfect” — Oak National Academy. A tutor still needs to check the material before it reaches a pupil.

  • Subject accuracy

    Check facts, methods, worked examples, terminology and answer keys.

  • Curriculum fit

    Check topic sequence, level, time available and exam relevance if the lesson is exam-focused.

  • Pupil suitability

    Adapt for confidence, prior knowledge, reading load, accessibility and emotional tone without adding identifiable data.

  • Bias and fairness

    Check examples, names, assumptions, cultural references and any hidden stereotype.

  • Safeguarding sensitivity

    Remove unsuitable activities, personal disclosures or advice that should not be handled by a lesson plan.

  • Final tutor judgement

    Keep what is useful, correct what is wrong, and adapt the lesson to the pupil you actually teach.

Secure the accounts and tools you use

Confidentiality is not only about what you type. It also depends on account security, tool settings and whether a third-party service is suitable for professional use. The NCSC says passkeys “should generally be used” where supported — NCSC.

  • Use stronger sign-in methods

    Use passkeys where supported, and use two-step verification where passkeys are not available.

  • Use a password manager

    Use strong, unique passwords for professional accounts and manage them securely.

  • Check data terms before any pupil data is involved

    Look for clear information on retention, use of inputs for training, sharing, administrator controls and deletion.

  • Do not rely on marketing claims alone

    A tool that sounds educational is not automatically suitable for personal pupil data.

  • Separate professional access where practical

    Avoid shared family devices or shared accounts for professional lesson preparation.

Message to check local AI rules

Suggested wording before using AI for professional lesson preparation

When this applies

Use before relying on an AI tool for professional lesson preparation where another organisation may set local rules, specified tools or additional restrictions.

Suggested wording

Hello, I am planning to use AI only for generic lesson preparation, such as lesson structures, activity ideas and explanations. I will not include identifiable pupil information, reports, SEND or health details, safeguarding-adjacent information, parent messages or school records. Are there any local rules, specified tools or additional restrictions I should follow?

Why this helps

It states the safe boundary clearly and invites stricter local requirements without claiming that any particular tool is permitted.

Sources and review note

This guide uses official UK regulator and security guidance first. Some ICO AI and data-protection pages are marked as under review following wider UK data-law changes, and AI provider terms can change, so avoid tool-specific confidentiality claims unless they are checked close to publication.

  • ICO: personal information guide

    Personal data, direct and indirect identification, anonymisation caution.

    Open source

  • ICO: what is personal data?

    Personal data, special category data, pseudonymisation and anonymisation.

    Open source

  • ICO: lawfulness in AI

    Purpose, lawful basis, documentation and privacy information.

    Open source

  • ICO: security and data minimisation in AI

    Data minimisation, AI security risks and third-party due diligence.

    Open source

  • ICO: AI accuracy guidance

    Predictions, inferences and output checking.

    Open source

  • NCSC: passkeys are the future

    Passkeys and two-step verification fallback.

    Open source

  • NCSC: password managers and passkeys

    Password managers and secure account habits.

    Open source

  • Oak National Academy: AI experiments

    Education AI examples and caution that tools still need human review.

    Open source

  • GOV.UK: Data (Use and Access) Act factsheet

    Freshness context for UK data-protection changes.

    Open source

Related guidance

More guidance from this section

More guidance from this part of the Ed Centre that may help with the same decision, stage or next step.

Support and clarity

Frequently asked questions

Straight answers to the questions people ask most often.

Can tutors use AI for lesson planning?

Yes, for generic or genuinely anonymous planning tasks such as lesson structures, quiz ideas, explanations and activity variations. Keep identifiable pupil data out of everyday AI use unless there is a lawful, documented and secure process. Treat the output as a draft that still needs tutor review.

Can I paste a pupil's report into AI if I remove their name?

Usually no. Removing a name does not automatically make the information anonymous. If the remaining details can still identify the pupil directly or indirectly, it remains personal data; the ICO also says pseudonymised information remains personal data.

What information should tutors keep out of AI lesson-planning tools?

Keep out names, contact details, school identifiers, reports, emails, behaviour notes and marked work tied to a real child. Treat SEND, health, wellbeing, family and safeguarding-adjacent details with heightened caution, and avoid combinations of details that could identify a pupil without a name.

Can AI help tutors plan for SEND learners?

Yes, AI can help with broad teaching supports such as chunked instructions, multisensory activities, lower-reading-load explanations and scaffolded practice. The safer approach is to describe the teaching need generally rather than entering a real pupil diagnosis, EHCP extract or health detail.

How should tutors check an AI-generated lesson before using it?

Check subject accuracy, curriculum fit, pupil suitability, bias, accessibility and safeguarding sensitivity. Keep what is useful, correct what is wrong, and adapt the material using professional judgement before teaching it.

What changes if identifiable pupil data is ever processed with AI?

The task is no longer ordinary generic planning help. The ICO says purpose, lawful basis, documentation and privacy information should be addressed before processing starts; special category data needs extra care and an additional condition. Higher-risk use may also need stronger governance and a DPIA.

Should tutors use a specific AI tool for lesson planning?

This guide does not recommend a named tool unless its current privacy, retention, training and account-security terms have been checked. For ordinary planning, focus first on generic inputs, secure accounts and human checking, whatever tool is used.

Sources and references

Sources and references

Official guidance

Other sources

  • 1.
    Oak National Academy: Oak AI Experiments

    Oak National Academy · No visible last-updated date on page · Accessed

    Used as an education-specific AI example and for the reminder that AI materials still need human review.