AI can save a teacher real time on planning, adapting materials, practice questions and routine emails. It cannot know your class, and it can be confidently wrong, so every output needs your judgement before a student sees it. Pupils’ personal data must stay out of it.
This page shows what to use it for, with starter prompts, and what else to keep out of it.
- Good for: first drafts of lesson plans, the same text at three reading levels, quiz questions, marking criteria and parent emails.
- Bad at: knowing your pupils, getting every fact right, and judging what a child actually understands.
- Privacy first: keep names and personal details out. England’s Department for Education recommends personal data is not put into these tools at all.
- Integrity: AI detectors are not reliable proof against a student. Drafts, class writing and a short conversation tell you more.
- The rule of thumb: treat AI like a keen trainee teacher. Useful drafts, but you sign off everything.
What AI is good at in teaching, and what it is bad at
The plain point: AI is a fast drafter, not a colleague who knows your room. It is strongest on the work around teaching, not the teaching itself.
A chatbot (a tool like ChatGPT, Gemini or Claude that answers in plain words) writes by predicting likely words. It does not look facts up unless it has a search tool. So it can produce a tidy worksheet with a wrong date in it. Why does AI hallucinate? explains the reason simply.
The comparison: think of a keen trainee teacher who has read every textbook but never met your class. They can draft a lesson in minutes. They cannot tell you that Year 8 struggled with fractions last week. You still plan, check and decide.
England’s education department lists the same kinds of tasks, such as creating resources, planning lessons and admin.1 It also says every output needs checking.1
| Task | How AI does | What you still do |
|---|---|---|
| Drafting | Good at: lesson outlines, worksheets, starter activities, emails. | You do: fit it to your class and check every fact. |
| Adapting | Good at: rewriting a text at an easier or harder level. | You do: check nothing important was lost or changed. |
| Facts | Weak at: dates, quotes, figures and sources. It can invent them. | You do: check against a trusted source. |
| Judging pupils | Bad at: knowing what a child understands or needs. | You do: all of it. This is the job. |
Hand AI the drafting. Keep the knowing: your pupils, your facts, your final call.
Six tasks to try, with starter prompts
The plain point: a good prompt (the request you type) gives the year group, the topic, the time and the format. Copy a box, change the parts in [brackets], and read the result as a draft.
The comparison: it is like briefing a supply teacher. “Do some maths” gets you a mess. “Forty minutes on equivalent fractions, Year 7, pizza examples” gets you a lesson.
1. A lesson plan outline.
2. The same text at three reading levels. This is one of the most useful jobs for mixed-ability classes.
3. Practice and quiz questions.
4. Marking criteria. A clear rubric (a table showing what a weak, good and excellent answer looks like) saves time when you mark.
5. Feedback ideas on anonymous work. Remove the name and anything that identifies the pupil first. You write the final comment.
6. Routine emails and admin.
For more on writing requests, see how to write AI prompts (Prompting 101) and how to ask AI well.
Give it the age, the topic, the time and the format, and ask it to list the facts you should check.
Student data: what never goes in
The plain point: do not type pupils’ names, grades, behaviour notes, health details or anything that could identify a child into a general AI tool.
The comparison: treat a chat like a postcard, not a sealed letter. It usually arrives safely, but people handling it could read it. You would not write a pupil’s support plan on a postcard.
England’s Department for Education is direct about this: “It is recommended that personal data is not used in generative AI tools.”1 It also covers pupils’ own work. Schools must not let it be used to train an AI (teach the next version of the model) without permission.1
That matters because many personal accounts use chats for training unless you switch it off. Business and education plans often differ. What happens to what you type explains the four separate questions: is it kept, is it used for training, can a person read it, and was it yours to share?
- Use your school’s approved tool and account, if it has one. Ask what the contract says about training.
- Anonymise. “Pupil A” and no details that point to one child.
- Never paste safeguarding notes, medical information or special needs records.
Your consent is not the one that counts. The data belongs to the child and their family, so the safe default is that it stays out.
Academic integrity without the guesswork
The plain point: AI detectors (tools that claim to spot AI-written text) are not reliable enough to accuse a student. Assessing the process works better.
In one study, seven detectors wrongly flagged most essays by non-native English writers as AI-written.2 The tools measure how predictable writing is, and simpler vocabulary reads as predictable. AI in schools covers the study and what it means.
The comparison: a detector is like a smoke alarm that goes off when you make toast. You would not call the fire brigade on that alone. You would go and look.
- Make the process visible: drafts, version history, some writing done in class.
- Talk to the student. Someone who wrote the work can explain their choices.
- Say what is allowed. For example: AI for ideas and grammar checks, but not for writing the answer.
- Ask for a declaration. Students who may say how they used it usually will.
In the UK, the Joint Council for Qualifications, which brings together the main qualification providers, sets rules for coursework.3 Its guidance says submitted work must be the student’s own, and any AI used must be named in the references, and should be dated.4 How to tell if something is written by AI lists signs worth a conversation, not a verdict.
A detector score is a reason to look, never proof. Drafts and a short conversation are fairer evidence.
What schools and governments advise
The plain point: official guidance agrees on three things. Protect pupils’ data, keep a teacher in charge, and check what the AI produces.
The comparison: it is like the rules for school trips. The bus is useful, but an adult stays responsible for every child on it.
UNESCO (worldwide). The United Nations’ education agency published guidance for schools and universities in 2023. It calls on governments to protect data privacy and to set an age limit for children talking to AI tools on their own.5 It also warns that schools are “largely unprepared to validate the tools”.6
England. The Department for Education says teachers must use professional judgement.1 For pupils, it asks for safeguards such as close supervision and filtered tools.1
Other countries. Rules differ by country, state and school. This page could not read the United States Department of Education’s guidance, because its site refused automated access on 1 Oct 2026. Check your own education department and your school’s AI policy. A team AI policy you can actually adopt is a starting point if your school has none.
Age rules. Each AI company also sets a minimum age, and some are 18. AI chatbot age limits and parental controls, app by app lists them in each company’s own words.
Guidance gives you cover when you follow it. Read your school’s policy once, and ask before using a new tool with pupils.
Which kinds of tools to look at
The plain point: start with whatever your school already provides. Then match the tool to the job, not the brand.
The comparison: it is like choosing between a whiteboard, a textbook and a calculator. Each is good at one thing.
- General chatbots for drafting plans, emails and materials. Choosing an AI model explains the differences without rankings.
- Tools built around your own sources. Google says notebooks in Gemini are open to “students of all ages, educators, and professionals”.7 Gemini Notebook, explained has the details.
- Education versions that guide rather than answer. Anthropic says its university product has a learning mode that “guides students’ reasoning process rather than providing answers”.8
- AI already inside your software. Many familiar apps added AI features. How AI got into your tools shows where to look in the settings.
Use the school’s approved tool first. For anything else, check the age rule and the training setting before you start.
Where to go next
Pick the line that sounds like you.
- “I suspect a student used AI.” Read AI in schools before you say anything.
- “I want to teach pupils to check AI.” Use how to check an AI citation and critical thinking.
- “I need to explain AI to my class.” Try how to explain AI to someone else and AI explained for complete beginners.
- “Parents are asking me about it.” Point them to AI for parents.
- “What should my students know?” Share AI for students and when not to use AI.
- “I want the safety basics.” Read how to use AI safely.
Questions people ask
How can teachers use AI to save time?
Use it for first drafts of lesson plans, the same text at different reading levels, quiz questions, marking criteria and routine parent emails. England’s education department lists the same kinds of tasks, and says every output needs checking.1
Can teachers put pupils’ names into an AI chatbot?
No. Keep names, grades, behaviour notes, health details and anything that could identify a child out of general AI tools. England’s Department for Education recommends that personal data is not used in generative AI tools.1
Are AI detectors reliable enough to accuse a student?
No. In one study, detectors wrongly flagged most essays by non-native English writers as AI-written.2 Drafts, version history, some writing done in class and a short conversation with the student are fairer evidence.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What does the page say about typing pupils’ names, grades or health details into a general AI tool?
- It is fine if you delete the chat afterwards
- Keep them out: the safe default is that pupil data stays out
- It is fine as long as the tool is popular
Show the answer
Answer: Keep them out: the safe default is that pupil data stays out. The page says the data belongs to the child and their family, so names, grades, behaviour notes and health details stay out.
2. An AI detector says a student’s essay was written by AI. What does the page suggest?
- Treat it as proof and give a zero
- Ignore it and say nothing at all
- Treat it as a reason to look, using drafts and a short conversation
Show the answer
Answer: Treat it as a reason to look, using drafts and a short conversation. The page says detectors are not reliable enough to accuse a student; drafts and talking to the student are fairer evidence.
3. What does a good teaching prompt include, according to the page?
- The year group, the topic, the time and the format
- Only the topic, so the AI has freedom
- The pupils’ names, so it feels personal
Show the answer
Answer: The year group, the topic, the time and the format. The page says to give the age, the topic, the time and the format, and to ask it to list facts you should check.
Related on this site: How to Cite ChatGPT and AI (APA, MLA, Chicago) · How to Learn a Language With AI · Manim Skills for Claude Code
- 1Department for Education (England), Generative artificial intelligence (AI) in education, updated 12 Aug 2025, read at source 1 Oct 2026“creating educational resources, lesson and curriculum planning, tailored feedback and revision activities, administrative tasks, supporting personalised learning”. ↑a“Any content produced requires critical judgement to check for appropriateness and accuracy.” ↑b“Personal data must be protected in accordance with data protection legislation. It is recommended that personal data is not used in generative AI tools.” ↑c“Schools and colleges must not allow or cause students’ original work to be used to train generative AI models unless they have permission, or an exception to copyright applies.” ↑d“Teachers, leaders and staff must use their professional judgement when using these tools.” ↑e“Pupils should only be using generative AI in education settings with appropriate safeguards in place, such as close supervision and the use of tools with safety and filtering and monitoring features.” ↑f“creating educational resources, lesson and curriculum planning, tailored feedback and revision activities, administrative tasks, supporting personalised learning”. ↑g“Personal data must be protected in accordance with data protection legislation. It is recommended that personal data is not used in generative AI tools.” ↑h
- 2Liang et al., Patterns 4(7), July 2023, read at source 16 Sep 2026“average false-positive rate: 61.3%”. ↑a“average false-positive rate: 61.3%”. ↑b
- 3JCQ, About us, read at source 1 Oct 2026“The Joint Council for Qualifications (JCQ) is a membership organisation bringing together the UK’s leading qualification providers.” ↑
- 4JCQ, Information for candidates: Coursework assessments, effective from 1 Sep 2026, read at source 12 Nov 2026 (replaces the JCQ AI guidance page this line cited until 12 Nov 2026, which no longer loads): “Where AI has been used, your reference must show the name of the AI tool used and should show the date the content was generated.”↑
- 5UNESCO, Guidance for generative AI in education and research, 7 Sep 2023, last updated 16 Jan 2026, read at source 1 Oct 2026“mandating the protection of data privacy” and “setting an age limit for the independent conversations with GenAI platforms”. ↑
- 6UNESCO, Guidance for generative AI in education and research, 7 Sep 2023, last updated 16 Jan 2026, read at source 1 Oct 2026.↑
- 7Google Workspace Updates, Notebooks in Gemini: a dedicated workspace for focused, organized work, now for schools and organizations, 17 Sep 2026, re-read directly at source 13 Oct 2026.↑
- 8Anthropic, Introducing Claude for Education, 2 Apr 2025, read at source 1 Oct 2026.↑
JCQ, AI Use in Assessments, 30 Apr 2025, and About us, both read at source 1 Oct 2026.
Liang et al., Patterns 4(7), July 2023, read at source 16 Sep 2026, as recorded on AI in schools.
Google Workspace Updates, Notebooks in Gemini, 17 Sep 2026, read in a browser 22 Sep 2026, as recorded on Gemini Notebook, explained, and re-read directly at source 13 Oct 2026.
The comparisons on this page (trainee teacher, supply-teacher brief, postcard, smoke alarm, school trip, whiteboard and calculator) are this site’s own ways of explaining, not claims from these sources.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK