AI Lesson Planning vs AI Tutoring: A UK Schools Guide

A multi-academy trust that says it is "looking at AI for teaching" is usually asking two different questions at once. One is whether a teacher should use AI to plan a lesson, build a quiz or adapt a worksheet. The other is whether a pupil should sit in front of an AI tutor that questions them, explains and adapts. The first tool serves an adult professional who checks everything before it reaches a classroom. The second serves a child directly, in real time, with nobody reading each reply before it lands.

The market blurs the two, and so do most "AI teacher" lists. The Department for Education does not. Its generative AI product safety standards ask every supplier to state which educational use case a product serves, and several of its sections apply only to learner-facing products. The evidence base also differs sharply: lesson planning has a randomised trial behind it, while the DfE's own guidance says evidence on pupils using generative AI themselves is still emerging.

This guide maps the AI teacher platform landscape in the UK by job, not by brand, so a trust education lead can decide which category to buy first. It deliberately cedes three things to neighbouring pages. The section-by-section safeguarding review of an AI tutor belongs to our AI tutor safeguarding guide for schools and MATs. The wider buying process for schools, universities and L&D teams belongs to the AI teacher buyer's guide. The product-by-product comparison with Khanmigo belongs to our Khanmigo alternative guide.

This guide is shipped by the team behind Dilr Academy, an AI tutor that builds interactive, multilingual courses on demand with Socratic questioning and mastery tracking. Or see DATS, our five-stage AI consulting system for placing AI inside regulated institutions.

What is the difference between AI lesson planning and AI tutoring?

AI lesson planning tools are teacher-facing: a teacher asks for a lesson plan, slides, a quiz or a worksheet, then edits the output before any pupil sees it. AI tutoring tools are learner-facing: a pupil works directly with the system, which asks questions, explains and adapts in the moment. The difference is who reads the output first, an adult professional or a child, and that changes the risk, the evidence and which DfE standards apply.

A simple test sorts almost any product a trust will be shown. Ask who types into the box. If the answer is a teacher, and the output is a draft the teacher owns, the tool is a planning assistant, however clever it is. If the answer is a pupil, and the system replies to that pupil without a teacher in between, it is a tutor or a digital assistant, however modestly it is marketed.

The test matters because the two categories fail in different ways. A planning tool that gets a fact wrong produces a flawed worksheet that a teacher can catch at the photocopier. A tutor that gets a fact wrong, or simply hands over the answer, does so inside a live exchange with a learner. That is why the questions a trust's AI operating model has to answer are different for each, and why it pays to decide which job you are buying for before you look at any brand.

How does the DfE classify AI tools for schools?

The DfE generative AI product safety standards, published in January 2025 and updated on 19 January 2026, ask developers to state their product's intended educational use cases from an ordered list of eight. Content creation and delivery, which covers tools that generate lesson plans and presentations, sits first. Digital assistant, which covers personal tutors and chatbots that guide learners, sits fourth. Several later sections, such as filtering, are stated as relevant to learner-facing products.

The eight use cases, in the DfE's own order, are: content creation and delivery; personalised learning and accessibility; assessment and analytics; digital assistant; research and writing aid; learner engagement and interaction; administrative and management; and other. A single product can select more than one, and the standards say the intended use case "should be clear to all users, educators and purchasers".

That list is the most useful map of the landscape a school has, because it is the DfE's own category line rather than a vendor's. The DfE does not name planners or tutors itself; in our reading, a planning assistant falls under use case 1, and sometimes 7 for administrative work, while an AI tutor falls under use case 4, and often 2 and 6 as well. The standards then state that filtering and monitoring are relevant to learner-facing products, including use cases 2, 4, 5 and 6, while sections such as design and testing, governance and manipulation apply to learner-facing and teacher-facing products alike.

Two points follow for a buyer. First, the standards are mainly intended for edtech developers and suppliers, so the expectation to evidence them sits with the vendor; a school uses them to judge what it is shown. Second, the standards set a tone that applies to every category: "Suppliers should not exaggerate the impact or capabilities of their tools." We read the full set section by section in our safeguarding review guide, so this page uses them only as the map.

What does the evidence say about AI lesson planning tools?

AI lesson planning has a randomised trial behind it in England, while the DfE says evidence on pupils using generative AI themselves is still emerging. An Education Endowment Foundation trial published on 12 December 2024 found that teachers using ChatGPT alongside a guide cut Year 7 and 8 science lesson and resource planning time by 31%, with no noticeable difference in quality. The finding is narrow: one subject, two year groups, one tool used with a written guide.

The EEF trial, evaluated independently by the National Foundation for Educational Research, involved 259 teachers in 68 secondary schools across England. Of those, 129 teachers in 34 schools were allocated to use ChatGPT with a guide on effective implementation. Their weekly Year 7 and 8 planning time averaged 56.2 minutes against 81.5 minutes in the comparison group, according to the EEF project page, a saving of 25.3 minutes a week. An independent panel of teachers found no noticeable difference in quality, though the EEF notes that judgement rested on a limited sample of resources.

Read the denominator carefully before quoting it to a board. The 31% is a share of Year 7 and 8 science planning time, not of a teacher's whole week, and the trial tested ChatGPT plus a guide, not any specific schools product. It is good evidence that a teacher-facing planner, used with structure, saves time without an obvious quality cost. It is not evidence about AI tutors at all.

The demand side is real but easing. In the DfE's Working lives of teachers and leaders wave 4 report, published in November 2025, 41% of teachers and middle leaders felt they spent too much time on individual lesson planning, the lowest figure in the series, against 71% for general administrative work. Full-time teachers averaged 50.1 hours a week in 2025, and full-time primary teachers 51.4. For a trust weighing where AI pays, our education hub maps the wider workload picture across admin, attendance and parent contact.

What can a teacher-facing planner like Aila do, and what can it not?

Oak National Academy's Aila is the clearest UK example of a teacher-facing planner. Oak describes it as an AI lesson assistant built for teachers that drafts lessons and resources from Oak's national curriculum-aligned content, and a lesson created with it includes a lesson plan, a slide deck, two quizzes and a worksheet. Oak's own page also states that Aila cannot create images or model diagrams.

On Oak's AI experiments page, Aila is described as working step by step with the teacher, "with you in control every step of the way", and every output is editable and downloadable. Oak adds that "You stay in control throughout, reviewing and adapting everything." That is what the planning category should look like: the teacher stays the author, the AI drafts, and the teacher decides what is used with a class.

Oak is also candid about limits. Its page says, at the time of writing, that Aila "can't create images or model diagrams, though it can suggest where to find them", that sample lessons do not currently include images, and that, like all AI tools, it is stronger in some subjects than others. For subjects where a diagram carries the idea, such as a circuit, a cell or a graph of a function, a teacher using a planner that cannot create diagrams still has to source or draw the visual themselves.

General-purpose assistants sit in the same category when a teacher uses them to plan. ChatGPT in the EEF trial was used for creating questions and quizzes, generating activity ideas and tailoring materials to groups of pupils. For a product-by-product view of Khanmigo, see our Khanmigo alternative guide. The common thread is that the teacher is the user and the editor.

What does a learner-facing AI tutor add?

A learner-facing AI tutor adds a live exchange with the pupil: it questions, checks reasoning, explains and adapts the next step to what the learner has shown. Dilr Academy is one example, an AI tutor that builds interactive, multilingual courses on demand with live diagrams, animations, simulations and Socratic questioning. The government announced a programme on 16 April 2026 to test AI tutoring tools in schools.

The government's AI tutoring programme, run by DSIT and the DfE, invited bids for safe, personalised tutoring tools for disadvantaged pupils. Up to 8 companies were to begin testing tools in schools from summer 2026 under teacher supervision, with the aim of making successful tools available nationally from 2027 and the potential to support up to 450,000 pupils a year. The announcement says every tool must meet the DfE's Generative AI Product Safety Standards, and that DSIT's Incubator for AI is developing national benchmarks with teachers. The government frames the gap in terms of cost: it says private tutoring is out of reach for many families despite evidence it can accelerate learning by up to 5 months.

Dilr Academy's live product page describes what the tutoring side looks like in practice. It asks before it tells and moves on only once the learner has understood. Its mastery tracking adapts to how the learner learns, with difficulty tuned by knowledge tracing. It renders live diagrams and charts inline, using Mermaid, D3 and tree visualisations, plus storyboard animations for hard ideas, and it is multilingual end to end, including right-to-left scripts. That visual layer is the clearest functional contrast with a planner that cannot create diagrams: in Dilr Academy the diagram is rendered inside the course as the learner works.

Be precise about what that does and does not mean. Dilr Academy is a learner-facing course builder with a live catalogue of 12 interactive courses across six or more subjects. Its live page describes a tutor for learners, not a lesson planning tool for teachers, and we do not claim here that its courses map to the national curriculum or meet any section of the DfE standards. The diagnostic logic behind how we would place it in a trust is the same one that drives our AI placement diagnostic, which starts from the job and the user, not the tool.

What changes when an AI tool faces pupils rather than teachers?

When an AI tool faces pupils, the DfE standards add sections on filtering, monitoring and cognitive development that are written for learner-facing products, on top of the school's existing safeguarding duties. A teacher-facing planner still needs data protection, security and governance checks, but its output passes through a professional first. A pupil-facing tutor has no such buffer, so the bar for evidence rises.

The DfE's guidance on generative AI in education, updated on 12 August 2025, leaves the choice of use case to each setting, as long as it complies with wider statutory obligations such as keeping children safe in education. It gives a plain example of how narrow that choice can be: "For example, schools and colleges may choose to only use AI tools with teachers, or only on administrative tasks." Others, it says, may use AI with students only in particular subjects, year groups or key stages, and pupils should only use generative AI in education settings with appropriate safeguards, such as close supervision and tools with filtering and monitoring.

That sentence describes a legitimate strategy, not a timid one. A trust can start with teacher-facing planning, where the EEF evidence is strongest, and move to pupil-facing tutoring later and narrowly. The same guidance is candid that "Evidence is still emerging on the benefits and risks of pupils and students using generative AI themselves." Read together, the two lines suggest a sequence rather than a single purchase.

What the learner-facing sections actually ask of a supplier, filtering that holds through a whole conversation, activity logging and supervisor alerts, progressive disclosure instead of full answers, is the subject of our AI tutor safeguarding review, and we do not repeat it here. The one practical point for this page is that those learner-facing sections apply only on the tutoring side of the line. A planner is still reviewed against the sections that apply to both categories, such as security, privacy, governance and manipulation, and our DATS methodology page explains how we approach AI governance more widely.

Should a school buy an AI planner or an AI tutor first?

A school or trust should usually buy for its most pressing job first, and where that job is teacher workload, it points to a teacher-facing planner. A planner has the stronger evidence, the lighter safeguarding load and a professional in the loop. An AI tutor is the right first purchase when the problem is pupil practice, catch-up or access to one-to-one support, and when the trust can run a supervised pilot properly.

The decision is easier if a trust works through it in order rather than starting from a vendor demo. The flow below is the sequence we recommend; it is our framework, not a regulatory requirement.

Naming the job is the step most often skipped. "AI for teaching" is not a job; "cut the time Year 7 science teachers spend building retrieval quizzes" is. Once the job is named, the user usually follows, and the DfE use case follows from the user. Only then does a vendor comparison mean anything, because a planner and a tutor are not competing for the same budget line.

The pilot step differs by branch. A planning pilot can be measured the way the EEF measured it: time spent and the quality of resources. A tutoring pilot needs a narrower scope, named year groups, supervised sessions and an agreed point at which the trust reviews logs and outcomes before widening access. Our AI execution office runs that kind of review cadence for institutions that want a senior team to hold it, and the operating model work defines who owns each decision once the pilot ends.

Many trusts will end up with both categories, and that is fine. The mistake is buying one while believing it is the other, for example adopting a pupil-facing chatbot as a "planning aid" and discovering later that the safeguarding review was never done. A placement diagnostic is designed to catch exactly that mismatch before procurement starts.

What is the best AI tool for lesson planning or tutoring in 2026?

There is no single best AI tool for both jobs in 2026, because lesson planning and tutoring serve different users. For teacher workload, a teacher-facing planner such as Oak's Aila, or a general assistant used with a guide as in the EEF trial, is the stronger choice. For learner practice with visual explanation, a learner-facing tutor such as Dilr Academy fits better. The right verdict depends on the job named first.

Use five criteria to make the call for your setting. First, the user: who types into the box. Second, the evidence: whether there is independent evidence for this category of use, and how narrow it is. Third, the DfE use case the supplier declares, and whether it matches how you will actually use the tool. Fourth, curriculum fit: whether you need content mapped to the national curriculum, which is where Oak's Aila draws its material from. Fifth, the visual layer: whether the subject depends on diagrams, which some planners cannot produce.

On those criteria, the honest concession is clear. If your bottleneck is teacher workload and curriculum-aligned resources, a teacher-facing planner wins outright, and Dilr Academy is not the tool for that job: its live page describes a tutor for learners, not a planning tool for teachers. Our best AI education tool guide, linked below, compares a wider field of AI education tools. Dilr Academy's case is narrower: structured, multilingual courses built on demand, with Socratic questioning and live diagrams, for learners who need to understand an idea rather than retrieve an answer.

For a trust running a structured comparison, the AI teacher buyer's guide sets out how procurement differs between schools, universities and L&D teams, and what Dilr Academy is explains the tutoring mechanics in more depth than this page needs to.

Frequently asked questions

Can one AI tool do both lesson planning and tutoring?

Some AI products do offer both a teacher-facing planning side and a learner-facing tutor, and the DfE standards let a supplier declare more than one educational use case. A school should still treat the two sides as separate decisions, because the pupil-facing side carries filtering, monitoring and cognitive development expectations that the planning side does not. Evaluate each side against its own job.

In practice that means two sets of questions in one procurement, and sometimes two different answers. A trust might adopt the teacher tools of a product while declining its pupil tutor until a supervised pilot has run. That is consistent with the DfE guidance, which explicitly allows a setting to use AI only with teachers. If the trust runs front-office automation as well, keep that on its own track too: attendance and parent calls are an administrative use case with their own data questions.

Does Dilr Academy plan lessons for teachers?

No. Dilr Academy's live page describes a learner-facing AI tutor, not a teacher-facing planning assistant. It builds interactive, multilingual courses on demand, with Socratic questioning, mastery tracking and live diagrams, for the person who is learning. A teacher who wants help drafting lesson plans, slides or quizzes from curriculum-aligned content is better served by a planning tool built for that job, such as Oak's Aila.

What it offers instead is a structured course: a real syllabus of modules, topics and stated outcomes, with a one-to-one AI tutor that adapts the next step to the learner, in the learner's own language. Beyond schools, our AI solutions page sets out how we place AI across a wider estate. If you are unsure which side of the line a use belongs on, talk to us before you shortlist.

Want to go further? Book an AI placement diagnostic to sort planners from tutors, compare the field in our AI education tools guide, read about our approach to placing AI inside institutions, or browse more AI strategy guides.

Written by the Dilr.ai engineering team, practitioners who ship enterprise AI in production. Follow us on LinkedIn for shipping notes, or subscribe via the RSS feed.