AI Lesson Planning for Teachers: A 90-Minute Weekly Workflow That Stays Yours

Published on: November 5, 2026 • Last Updated: November 6, 2026

Short answer: Use AI to draft a lesson plan in about 90 minutes a week, then spend the saved time on the parts only a teacher can do — checking the level, fixing the examples, and planning how you will know whether the class understood it. The draft is the starting point, not the lesson.

Most lesson-planning tools promise the same thing: type in a topic and get a finished plan. That promise is exactly where the trouble starts. A plan that looks complete on screen can still be pitched a year too high, built around an example your class will not recognise, or full of objectives that sound precise and mean very little. Teachers do not need a faster way to produce that kind of plan. They need a workflow that keeps the teacher in charge of the decisions that matter.

This guide sets out one such workflow. It works for a single lesson or for a week of lessons, it uses one general tool and one source-grounded tool rather than a shelf of apps, and it ends with a check that takes ten minutes and catches most of the errors AI drafts tend to contain.

Start from the objective, not the tool

Before you open any AI tool, write three sentences on paper or in a blank document:

• By the end of this lesson, students will be able to do what, exactly?
• What will they produce, say, or solve that proves it?
• What do they already know that this lesson builds on?

That sounds like extra work, but it is the step that makes everything after it useful. An AI draft written without an objective will invent one, and the invented objective is usually the vaguest possible version: “understand photosynthesis” rather than “explain, in their own words, why a plant kept in the dark stops growing”. When you hand the tool a sharp objective, the draft comes back sharper too, and you can judge it against something real instead of admiring how fluent it sounds.

Keep the objective in the prompt for every step that follows. If a later draft drifts away from it, you have a clear reason to cut.

The 90-minute weekly workflow

The workflow below assumes one lesson, but it scales: repeat the draft step for each lesson in the week, then do the checks in one sitting. The times are a guide, not a promise — the point is the order, because each step only works if the previous one is done.

Step 1: Draft the skeleton in 20 minutes

Ask a general AI assistant for a lesson skeleton, not a finished plan. Give it the year group, the topic, the objective from the previous section, the lesson length, and one constraint from your classroom — for example, “no group work longer than ten minutes” or “half the class is learning English as an additional language”. Ask for a hook, two or three main activities, a check for understanding, and one homework or follow-up task.

Do not ask for a full script at this stage. A skeleton leaves room for your judgement, and a weak activity is easier to spot in a short outline than inside five pages of polished text.

Here is a prompt shape you can reuse, filling in the bracketed parts:

Step 2: Ground the content in your own sources in 20 minutes

General assistants are fluent, but they are not grounded in your textbook, your scheme of work, or last week’s lesson. That is where a source-grounded tool earns its place. Upload the chapter, handout, or scheme document you are actually teaching from, and ask questions against that material only: which definitions does the chapter use, what examples does it give, and where does it warn students about a common confusion?

Google’s NotebookLM, renamed Gemini Notebook in July 2026, is built for exactly this job: it answers from the documents you upload and points back to the passage each answer came from. Our NotebookLM guide for students shows the same source-grounded approach step by step. Use it to check the draft skeleton against your source, not to generate a second, competing plan. Where the skeleton and your source disagree — on a definition, a sequence, or an example — trust the source, and note the disagreement so you do not reintroduce it later.

This step is also the one that protects you on accuracy. A fluent draft can state a wrong fact with complete confidence. A source-grounded check gives you a passage to open and read, which is a different kind of evidence from “the AI said so”.

Step 3: Differentiate in 15 minutes, not as an afterthought

Most AI drafts add a line at the end saying “differentiate as needed” and leave the actual work to you. Do the work explicitly, in three passes:

• Support: one version of the main task with a sentence starter, a worked first step, or a smaller set of items. This is for students who need a way in, not a different lesson.
• Core: the task as written, for most of the class.
• Stretch: one question that asks students to apply the idea somewhere new, explain why a plausible wrong answer is wrong, or connect this lesson to an earlier one.

For students learning English as an additional language, ask for the same task with key vocabulary defined in plain language and with a visual or diagram added — not a simplified topic. Simplifying the topic teaches less; clarifying the language teaches the same topic accessibly. That distinction is worth stating in the plan, because an assistant left to itself will often do the first when you asked for the second.

Step 4: Build the materials, don’t just describe them

A plan that says “worksheet” or “slides” is not finished. In this step, turn each activity into the thing students will actually see: the five questions, the diagram with labels blanked out, the exit ticket with three questions on it. Generate these from the checked skeleton, not from a fresh prompt, so the materials match the lesson you have already approved.

If you work in Google Workspace, this is where the built-in AI in Docs and Slides helps most: drafting a handout from your outline, or turning a set of points into a slide deck you then edit. See our Gemini in Google Workspace guide for how those tools fit a Docs, Sheets, and Slides workflow. Keep the deck short. A slide for every sentence in the plan is a sign the tool, not the teacher, decided the pace of the lesson.

Step 5: Check before you teach it in 10 minutes

Read the finished plan once, slowly, with this checklist in front of you:

• Facts: Is every definition, date, and example one you could defend from your source? If not, check it or cut it.
• Level: Would a student in this year group understand the vocabulary without you translating every sentence? If the language is a year ahead of the class, rewrite it now, not mid-lesson.
• Objective: Does the check for understanding actually test the objective you wrote at the start? A quiz on vocabulary does not prove students can explain a process.
• Time: Add up the timings honestly. AI drafts routinely plan 60 minutes of activity into a 45-minute lesson.
• Student data: Did you paste anything into a tool that you should not have — names, grades, or identifiable student work? If in doubt, remove it and use a made-up example instead.
• Policy: Does your school have a rule about AI use, marking, or homework that this plan touches? A plan that breaks school policy is not a time-saver, however good it looks.

If a plan fails the facts check or the objective check, fix it before you teach it. If it only fails on polish, teach it and fix the polish afterwards. Perfection in the document is not the goal; a lesson you can defend is.

Which tool for which job

You do not need a dozen tools. You need one tool for each job, and a clear reason for choosing it. The table below is a starting point, not a ranking — check each tool’s current free tier and eligibility on its own site before you commit, because teacher offers and prices change often.

Job	What to use it for	Watch out for
Drafting the skeleton	A general assistant such as ChatGPT, Claude, or Gemini, with your objective and constraints in the prompt	Fluent but ungrounded — always check facts against your source
Checking against your source	A source-grounded tool such as Gemini Notebook, working from your uploaded chapter or scheme	Only as good as the documents you upload
Worksheets and handouts	The same general assistant, or AI features in Docs, working from the checked skeleton	Materials that do not match the lesson if prompted separately
Slides and visuals	AI features in Slides, or a design tool with an education offer, for diagrams and short decks	Decks that set the pace instead of supporting it
Differentiation	The general assistant, asked for support, core, and stretch versions of the same task	Simplified topics offered in place of clarified language

Teacher-specific tools such as MagicSchool and Brisk Teaching are worth a look, but they do not replace the check step: a specialist tool can still invent a standard code or misjudge a reading level. The workflow above works with any of them, because the teacher’s checks sit outside the tool.

An illustrative example: one science lesson

The following example is illustrative, not a report of a real classroom. It shows how the workflow plays out for a single 45-minute lesson.

A teacher planning a Year 8 lesson on photosynthesis starts with the objective: “By the end of the lesson, students will be able to explain, in their own words, why a plant kept in the dark stops growing.” The skeleton draft comes back with a hook about a plant on a windowsill, a practical observation activity, and a written explanation task. The source check against the textbook chapter corrects the draft’s definition of a limiting factor and supplies the chapter’s own example, which the class has already met. Differentiation adds a sentence starter for the explanation (“A plant in the dark stops growing because…”) and a stretch question asking why a plant in dim light grows slowly rather than not at all. The exit ticket asks three questions: one recall, one explanation, and one prediction. Total planning time, including the check, is just over an hour — and the lesson that results is one the teacher can explain, defend, and adjust mid-lesson, because she built it rather than merely receiving it.

Where AI helps most, and where it misleads

AI is strongest at the first draft and the routine variation: turning one task into three levels, generating ten practice questions from a set of notes, or rephrasing an explanation for a different reading level. It is weakest at exactly the judgements teachers are trained to make: whether an example will land with this class, whether a question is fair for this age group, and whether an activity will take ten minutes or twenty-five.

Two failure modes deserve special attention. The first is invented precision: objectives, standard codes, and assessment criteria that look official and are not. If a draft cites a code or standard you do not recognise, verify it before you put it in a plan that a colleague or inspector might read. The second is quiet drift: a plan that starts on your objective and, through successive drafts and additions, ends up covering three topics shallowly instead of one topic well. The objective you wrote at the start is the antidote to both — return to it whenever the plan starts to feel impressive rather than teachable.

Build a plan bank, not a pile of one-off drafts

The real time saving from AI planning does not come in week one. It comes in week ten, when you reuse a checked plan instead of generating a new one. After each lesson, spend five minutes noting what worked: which hook landed, which activity ran long, which question confused the class. Save the corrected plan, not the first draft, and save the prompt that produced the useful version alongside it.

Over a term, this builds a small bank of plans you trust, matched to your classes. A plan you have taught, corrected, and annotated is yours in a way no first draft can be.

What to do this week

Pick one lesson you have to plan anyway. Write the objective in three sentences. Run the five steps above, with the ten-minute check at the end, and note how long each step actually took. The same objective-first habit sits behind our guide to using AI tools for studying and our weekly study planner that adjusts itself. If the draft saves you time without weakening the lesson, keep the workflow and add a second lesson next week. If it costs you more time in checking than it saves in drafting, the problem is usually the prompt, not the workflow: tighten the objective, narrow the ask, and try once more before deciding the tool is not for you.

Frequently asked questions

Will AI lesson plans all look the same?

Only if you teach the first draft. A plan that has been checked against your source, differentiated for your class, and corrected after teaching carries your decisions in it, and it will not look like anyone else’s.

Is it safe to upload my scheme of work or textbook chapter?

Check your school’s policy first, and never upload student names, grades, or identifiable student work. Uploading your own teaching materials is a different question from uploading student data, and the two should not be confused.

What if the AI draft gets a fact wrong?

Treat it as a normal drafting error, not a crisis: check the fact against your source, correct it, and note the correction so the error does not return in a reused plan. The check step exists because errors happen, not because they are rare.

Do I need a paid tool to do this well?

No. The workflow in this guide uses a general assistant and a source-grounded tool, both of which offer free access for teachers in various forms. Check current eligibility and limits on each vendor’s site before you rely on a particular offer.

Does the 90-minute workflow save time in the first week?

Not necessarily. The real time saving does not come in week one — it comes in week ten, when you reuse a checked plan instead of generating a new one. In the first week, note how long each step actually takes, and judge it honestly: if the draft saves you time without weakening the lesson, keep the workflow and add a second lesson next week. If checking costs you more time than drafting saves, the problem is usually the prompt, not the workflow — tighten the objective, narrow the ask, and try once more before deciding the tool is not for you.

Can I use this workflow for a whole week of lessons, not just one?

Yes. The workflow is written for one lesson, but it scales: repeat the draft step for each lesson in the week, then do the checks in one sitting. The times are a guide, not a promise — the point is the order, because each step only works if the previous one is done.