Never stare at a blank lesson plan again. This course shows you exactly how to use AI to plan faster, create better materials, and save 5–10 hours every single week — starting today.
Start the Course"The teachers winning right now aren't working harder. They're using AI as a junior assistant — and getting their lives back."
The average expat ESL teacher spends 8–12 hours a week on lesson planning and material creation. That's two full workdays — gone. Not on teaching. On admin.
This course doesn't just introduce you to AI tools. It gives you a complete, repeatable workflow. By the end you'll have a system that handles the boring parts of your job so you can focus on the part that actually matters: being in the room with your students.
Most teachers who try AI and give up broke one of these. Read them once. Remember them always.
Every module below contains the real thing — explanations, frameworks, worked examples, and copy-paste prompts that have been tested and proven. Click any module to open it. Read it. Apply it. Then move to the next.
Before you write a single prompt, you need to understand your tools. Not at a surface level — but deeply enough to know which tool to reach for, why, and what it's actually doing when you type. This module sets that foundation. It takes 20 minutes to read. It will change how you work forever.
When you type into ChatGPT or Claude, you're talking to a Large Language Model (LLM). Here's the simplest way to think about it: an LLM has read a significant portion of the internet, millions of books, and billions of documents. It learned the patterns of how language works across all of that. When you give it a prompt, it predicts — with extraordinary accuracy — what the ideal response looks like based on everything it has read.
This means two things for you as a teacher. First: it's remarkably good at generating language at specific levels (because it has read thousands of A1 texts, thousands of B2 texts, etc.). Second: it has no memory between sessions unless you give it context. Every conversation starts fresh. This is why your prompts need to include all the relevant context every time — level, age, topic, objective. The AI isn't guessing. It's pattern-matching from what you give it.
There are dozens of AI tools now. Most of them are irrelevant for ESL teaching. Here are the three that matter, what each one is best at, and exactly how to set them up.
How to use Claude effectively: Claude responds best to structured prompts. Use line breaks and categories. Tell it exactly what you want, in what order, in what format. Claude also has a "Projects" feature (free) where you can save a system prompt — a set of background instructions that apply to every conversation in that project. This is where you'll save your Class Context Block (covered in Module 6). Set up one Project per class and your context loads automatically every time.
One critical Claude behaviour to understand: if you give it vague instructions, it will make reasonable assumptions and produce something generic. If you give it precise instructions — CEFR level, age group, L1, learning objective, format — it produces something you can use almost unchanged. The gap between a generic Claude output and an excellent one is entirely in how you ask.
How to use ChatGPT effectively: ChatGPT's free tier now includes image generation through DALL-E. This is useful for creating visual vocabulary aids — describe what you want and it generates a usable image. For text tasks, ChatGPT responds well to direct, clear prompts. It tends to produce shorter, more bullet-pointed outputs than Claude — useful for quick activity ideas, less useful for complete lesson plans. Use it for speed; use Claude for depth.
A workflow that works well: brainstorm activity ideas in ChatGPT (fast, punchy), then take the best idea into Claude and expand it into a full lesson plan (precise, structured). These two tools complement each other perfectly.
How NotebookLM works: You create a "Notebook" (think of it as a workspace), upload your source documents as PDFs, and then chat with them. The AI's responses are grounded entirely in what you uploaded — it won't make things up from its training data. This makes it uniquely trustworthy for course book work.
The Audio Overview feature is genuinely remarkable for ESL teaching. Upload any document — a chapter, a vocabulary list, your own notes — click "Generate Audio Overview" and NotebookLM creates a 10–15 minute natural-sounding podcast discussion of the content. Two AI voices discuss the material conversationally. Hand this to students as a listening exercise or revision tool. It takes 2 minutes to create and replaces hours of work.
Here is what happens to 80% of teachers who try AI: they open ChatGPT, type "make me a lesson about food for my class," get something mediocre, conclude "AI isn't that useful for teaching," and never try again. This is entirely a prompting problem, not a tool problem.
The fix is understanding what the AI needs to produce something excellent. It needs the same information a supply teacher would need before walking into your classroom: who the students are, what level they're at, what topic you're covering, what you want them to be able to do by the end, how long you have, and any constraints (no printing, small class, mixed ability, etc.).
The difference in those two prompts is not intelligence — it's information. The second prompt took 20 extra seconds to write and saved 40 minutes of editing. That's the entire skill of working with AI: giving it enough context to do its job properly.
Open Claude and ChatGPT side by side. In both, run this exact prompt — but fill in the brackets with your real class details before sending:
Compare the outputs. Notice: Which one followed your format instructions better? Which vocabulary felt more level-appropriate? Which warmer felt more realistic for your classroom? Write one sentence about each. You now have an informed opinion about which tool to reach for and when.
This module gives you the most valuable thing in this entire course: a single prompt framework that generates a complete, structured, classroom-ready lesson plan for any topic, any grade, any level — in under 3 minutes. Learn it once. Use it every week for the rest of your teaching career.
Lesson planning from scratch has three main time costs. First: the blank page problem — starting with nothing and trying to shape it into a structured experience. Second: the "what activity next?" problem — deciding what comes after what, at what pace, for your specific class. Third: the "is this the right level?" problem — second-guessing whether your vocabulary, texts, and tasks are actually appropriate.
AI eliminates all three. You give it the context. It eliminates the blank page and makes the structural decisions. Your job becomes editing and personalising — not creating from zero. A plan that took 90 minutes now takes 10. That's not an exaggeration. That's what teachers consistently report after implementing this system.
A lesson plan is only as good as the information you feed into it. These 8 variables are what separate a generic, borderline-usable plan from one you can walk into the classroom with on Monday morning. Think of them as the briefing you'd give a substitute teacher who needs to deliver your lesson.
This is the prompt. It has been structured to give Claude exactly what it needs to produce a complete, lesson-plan-ready output. The brackets are your variables. Fill them in with your real class details before sending. Do not leave any bracket empty — each one is doing real work.
Pick a class you're teaching this week. Fill in every bracket of the Master Blueprint Prompt with real values from your actual class. Send it to Claude. Read the output once as a teacher — not as a prompter. Identify one thing you'd change. Edit it. Save the result. That plan took under 10 minutes. That's your new benchmark.
You have a lesson plan. Now you need the stuff — worksheets, reading texts, conversation cards, grammar drills, vocab activities. This module shows you how to generate any classroom material in under 2 minutes, at exactly the right level, on exactly the right topic. No more recycling old photocopies. No more paying for materials packs. No more searching online for 45 minutes to find a text that's almost right.
When you ask a colleague to make you a worksheet, you don't say "make me a worksheet." You say "can you make me a gap fill on past simple irregular verbs for my B1 teens — about 12 sentences, topics related to travel, answer key at the bottom?" That's exactly how you prompt AI. The more specific your request, the more usable the output.
The golden rule for materials creation: always end your prompt with "Include a complete answer key. Flag any item where more than one answer could reasonably be correct." This one sentence saves you the embarrassing moment in class where a student gives a perfectly valid answer that doesn't match your key.
This is the prompt format that generates a complete, print-ready worksheet — multiple exercise types, reading passage, and answer key — in a single request. It works because it specifies every component the AI needs to structure a coherent, level-appropriate activity sequence.
One of the biggest time wasters in ESL teaching is finding reading texts. You need something at exactly B1 level, on exactly your topic, that isn't too long, isn't too short, and doesn't have cultural blind spots. AI generates custom reading passages in under a minute — tailored exactly to your class's level and interests.
Conversation cards are one of the highest-value ESL materials because they facilitate genuine communication practice with minimal teacher input once distributed. They're also among the easiest materials for AI to generate well — because conversation questions are pure language, no formatting complexity required.
The key to great conversation cards is not just the question — it's the follow-up prompts. A bare question like "Do you use social media?" leads to one-word answers. A card that says "Do you use social media? → Why/Why not? → How has your use changed in the last year?" produces real conversation. Always prompt AI to include follow-up starters.
Pick your hardest class to make materials for — the one where you always spend the most time searching for the right resource. Use the Complete Worksheet Generator above with that class's real details. Run it. Check the answer key carefully. Edit anything that needs it. Print it and use it this week. Then calculate how long that took compared to your usual method for the same outcome.
Mixed-ability classes are the defining challenge of ESL teaching. You have students who are miles apart in level, and you're supposed to keep everyone engaged and progressing simultaneously. Before AI, the realistic options were: teach to the middle (bores the strong, loses the weak), or spend hours creating multiple versions of every material. Now there's a third option: run one prompt and get three versions in 90 seconds.
You've seen A1–C2 your whole teaching career, but let's make it concrete for AI prompting. When you tell AI a level, this is what should change in the output:
This is the most time-saving prompt in the entire course for teachers with mixed-ability classes. Take any existing lesson, worksheet, or activity. Paste it. Get back three differentiated versions — A, B, and C level — in a single response. Same topic. Same learning objective. Different difficulty.
The reason this works so well is that AI has seen thousands of texts at every CEFR level. It knows intuitively what B1 vocabulary looks like vs A2 vocabulary. You don't have to specify every individual change — you specify the level and let the AI apply everything that level implies.
Level adaptation is about language difficulty. Age adaptation is about everything else: activity types, attention span, topic relevance, vocabulary context, tone of instructions, cultural references. A B1 lesson for 8-year-olds looks nothing like a B1 lesson for 30-year-olds. Here's the prompt that handles the full adaptation automatically.
Take a lesson or worksheet you used this week. Paste it into the Level Fork prompt. Get back all three versions. Compare them side by side and notice exactly what changed between A2 and B2 — that gap is your practical understanding of the CEFR scale. Print all three versions and use them with your next mixed-ability class. See what happens.
Grading 30 essays on a Friday night is where teachers burn out. Not the teaching itself — the admin loop that follows it. This module gives you a complete AI feedback system that produces better, more consistent written feedback in a fraction of the time. And unlike your handwritten comments at 11pm, the AI feedback actually follows research-backed principles every single time.
Research on feedback in language learning is consistent on three things: feedback that focuses on too many errors at once leads to little improvement; feedback that starts with a genuine strength gets more engagement from learners; and feedback that tells students what to do next time (feed-forward) is more effective than feedback that only describes what went wrong last time.
Here's the honest problem: when you're marking 30 papers at 11pm on a Thursday, you're not following any of those principles. You're correcting everything. You're getting frustrated. The "well done!" at the top isn't specific. And you're not writing "next time, focus on..." because you're exhausted. AI doesn't get tired. Properly prompted, it follows all three principles every single time, for every single student.
Creating a rubric from scratch takes time and requires careful thinking about what distinguishes good from excellent. AI can generate a complete, specific rubric for any writing or speaking task in under a minute — with descriptors that are genuinely observable and specific, not vague placeholder phrases like "good vocabulary range."
Here's a workflow that turns your marking from a chore into a teaching opportunity. After marking your class's writing, you'll notice patterns — the same errors appearing across multiple students. Feed those patterns into AI and it gives you: the likely root cause, the most effective classroom activity to address it, and a ready-made exercise you can use in your next lesson.
Take 5 pieces of recent student writing. For each one, run the Written Feedback prompt with the actual student text pasted in. Add one personal touch to each output. Return the feedback and observe how students respond compared to your normal marking. Then list the 3 most common errors across those 5 papers and run the Class Error Pattern prompt — use the resulting Language Focus slot in your next lesson.
Most teachers use AI like a vending machine — put in a coin, get something out, forget it. The teachers who get 10x more value from the same tools are the ones who save what works. This module shows you how to build a prompt library that makes every week faster than the last — and gives you a plug-and-play context system so AI always knows your classes without you having to re-explain them.
A great prompt is an asset. It took you time to get right — to find the exact variables, the exact format instructions, the exact phrasing that gets you a usable output consistently. If you don't save it, you'll spend that time again next week, get a slightly different result, and wonder why last week's felt better.
A prompt library is simply a document where you store your best prompts, organised so you can find them fast. Nothing fancy. A Google Doc with sections. Notion database if you want to get organised. Even a .txt file. The storage system doesn't matter. What matters is that you save, tag, and reuse.
Here's the compounding effect in practice: Week 1, you build 5 prompts. Week 2, you use 3 from the library and build 2 new ones. Week 4, you're using 10 saved prompts and building almost none from scratch. By Month 2, you have a library of 30–40 prompts covering every class you teach. Planning a full week takes 30 minutes. This is the goal.
Every time you start a new Claude conversation, the AI has no memory of your classes. This means every session, you either give it context or it generates generic outputs. The solution is a Class Context Block — a short, structured paragraph about each class that you paste at the start of every session involving that class.
Once you have this habit, everything AI generates for that class becomes dramatically more relevant. It knows your students are teenagers who love K-pop. It knows you have 3 strong students who need extension work. It knows you can't print more than 1 page per student. It produces outputs calibrated to your actual reality.
Claude's Projects feature (available free) takes the Class Context Block one step further. Instead of pasting context at the start of every conversation, you save it once in the Project instructions — and it's automatically included in every conversation you start within that Project.
Here's how to set it up: In Claude, click "Projects" in the sidebar. Create a new project called the name of your class. In the "Project Instructions" section, paste your complete Class Context Block. Done. Now every conversation you start in that Project automatically has your class context loaded. You never have to paste it again.
Create one Project per class. This is the closest thing to an AI that actually knows your students — it won't forget between sessions, and every output it generates will be calibrated to your real classroom from the first message.
Right now — before Module 7 — do these three things: 1) Create a Google Doc called "My Prompt Library." Add sections: Planning / Materials / Feedback / Differentiation. Copy the best prompts from Modules 1–5 into the right sections. 2) Create a Class Context Block for your top 3 classes using the template above. 3) Create a Claude Project for each of those classes and paste the context block in as Project Instructions. You now have a functioning AI system for your teaching. Everything from here gets faster.
This is the module where things get genuinely impressive. NotebookLM is a free tool from Google that does something neither ChatGPT nor Claude can: it reads your actual documents — your course book, your syllabus, your own notes — and lets you interrogate them directly. No hallucination. No generic filler. Pure, cited answers from exactly what you uploaded. Combined with the tools you've already learned, this makes you one of the most efficient teachers in your school.
Go to notebooklm.google.com. Log in with any Google account. Click "New Notebook." You now have a workspace that can hold multiple source documents — PDFs, Google Docs, web URLs, pasted text. Upload your course book chapter, your school's syllabus, your own lesson notes, anything relevant to a unit you're teaching.
Now here's what makes it different from ChatGPT: when you ask a question, NotebookLM answers only from what you uploaded. It won't blend in outside knowledge. It cites the source with a direct quote so you can verify. This makes it extraordinarily useful for course book work, because you can ask things like "What vocabulary does Unit 5 introduce?" and get an accurate, cited answer from the actual book — not a guess.
NotebookLM lets you share notebooks with students directly. Here's a powerful setup: create a notebook per course unit, upload the chapter, vocabulary list, and any supplementary materials. Share the link with students. They can now ask the notebook questions — "What does 'phenomenon' mean?" "Can you quiz me on the Unit 5 vocabulary?" — and get answers grounded in the actual course material.
This creates a curriculum-aligned self-study tool that requires no teacher time to maintain after initial setup. Students who miss a class can catch up. Students preparing for tests can quiz themselves. Students who want to go deeper can explore. All within the safe boundaries of what you uploaded — no internet rabbit holes, no off-topic AI conversations.
The last use of AI in this course is for you, not your students. AI is an excellent reflective thinking partner. After a difficult lesson, a lesson observation, or a challenging class, you can use Claude to process what happened, identify the root cause, and find a practical solution — faster and more structured than journalling alone.
Go to notebooklm.google.com right now (it takes 2 minutes). Upload any PDF you have — a course book chapter, a unit from a textbook, even your own lesson notes. Run the 5 queries above. Then click "Generate Audio Overview" and listen to the first 3 minutes of what it creates. Think about one class where you could give students access to a notebook like this for self-study. That's your implementation plan for this week.
Everything in this course runs on free tiers. No subscriptions required to implement a full AI teaching workflow.
Every prompt in this course works for any class you teach. Here's how the system adapts across age groups and levels.
These are the 40+ field-tested prompts from across the entire course, organised by category. Every single one has been verified to produce classroom-ready outputs. Bookmark this section. This is your weekly reference.
This is the capstone. One Sunday session. Everything from this course applied to your real teaching week. Follow this exact sequence once — you'll never plan the old way again.