Lesson Planning in ESL: What AI Does Well and Where Teachers Should Draw the Line

Seventeen kids were staring at me while I tried to explain the difference between past simple and present perfect using three markers and a dry whiteboard. My newcomer students had glassy eyes, my intermediate kids were bored, and I had forty minutes left. That was the Tuesday I stopped pretending I could write three-tier differentiated ESL lesson plans by hand every single night. If you teach English language learners, you already know the grind. You need a text that an interrupted-formal-education teenager can read, comprehension questions that do not rely on cultural idioms they have never heard, and vocabulary work that actually sticks.

We started testing AI lesson planning for esl teachers this term to see what survives contact with a real classroom. The tools we looked at include MagicSchool AI, Eduaide.Ai, ChatGPT, Canva for Education, Synthesia, and Gamma. Some of them save hours. Others create more cleanup work than they remove. If you want a wider look at how these systems fit into different classrooms, check our guide on best AI tools for teachers 2026. Here is what actually works when you are planning for a room where five languages are spoken at once.

Where AI Saves Time in Language Planning

The best use of these tools is cutting down the mechanical prep of leveling texts and building scaffolds. When you need a news article rewritten so that a CEFR A2 learner can parse it without an expensive textbook, generative tools are genuinely fast. You paste in a paragraph about local weather and ask for simplified syntax and a glossary of five high-frequency verbs. That keeps you from spending your Sunday night digging through six different curricula for something appropriate.

MagicSchool AI is particularly strong here because its dashboard has purpose-built template libraries that require almost no prompting skill. You click into a generator, fill out a form for reading level and topic, and get a structured output. It is our top pick for parent-facing writing, such as translating complex school policy letters into accessible home languages. Its free tier covers core generators with monthly usage limits, while paid access runs around $12 per user per month. The main catch is that you still need to review everything for cultural nuance and local standards alignment.

Eduaide.Ai handles differentiation well, making it our choice for mixed-ability classes. It focuses on lesson, rubric, and worksheet generation with a clean, uncluttered interface. The free tier uses a daily generation cap, and the paid plan is around $10 per user per month. It does not have as many templates as MagicSchool, but its worksheet outputs require fewer edits before you send them to the copy machine.

The Prompt That Survived a Tuesday Lesson

Prompt: Act as an experienced ESL curriculum writer. Take this intermediate science paragraph about photosynthesis and generate three distinct group tasks for a mixed-ability class. Group A needs heavy sentence frames and vocabulary matching. Group B needs a structured graphic organizer with cause-and-effect headings. Group C needs a roleplay prompt where they argue from the perspective of different plant cells. Keep instructions under 50 words per task.

First draft output: The AI gave Group A a clean matching list, gave Group B a workable chart, and gave Group C a creative cell dialogue. However, it included a fifth vocabulary word in Group A that was far too advanced, and it forgot to include an extension task for early finishers in Group C.

Two edits that made it usable:

1. Swapped out the advanced vocabulary term in Group A for a high-frequency cognate from the glossary.
2. Added a one-sentence extension question to Group C asking students to write down their character's main defense mechanism.

Building Visuals and Decks Without Designer Fatigue

Text is only half the battle in an ESL classroom. If your students cannot see the vocabulary, they cannot anchor it. Canva for Education is effectively free for verified teachers, though account verification can sometimes take a few days. It gives you access to a massive template library with AI writing and image generation features. You can spin up vocabulary flashcards with clear illustrations in minutes, which is miles better than searching Google Images for royalty-free pictures of abstract concepts like frustration or ambiguity.

When you need to turn a rough outline into a slide deck for direct instruction, Gamma is the fastest route. It generates presentation decks and documents from a single prompt, offering modern-day layouts by default. The free tier relies on credit limits, and the paid option sits around $10 per user per month. Design control is limited compared to building from scratch, but it stops you from spending an hour aligning text boxes on a Monday morning. For a related workflow on handling student output, see our guide on AI essay feedback for history teachers.

Where Generative Tools Fall Down

Not everything works. If you ask an AI to generate a quiz from an authentic PDF article, check every single answer key twice. Numerical items and inference questions are the least reliable parts of the output. Generators love to sneak in ambiguous distractors that even native speakers trip over, which is the last thing you want when testing an ELL student's command of a grammatical structure.

Synthesia attempts to solve the speaking practice problem by turning scripts into presenter-led video with AI avatars. At around $29 per user per month with no free ongoing tier, the cost is hard to justify for a single classroom teacher. While its multilingual voice output is impressive, avatar video often feels impersonal for language learners who desperately need human interaction, eye contact, and authentic conversational pauses. Do not use student likenesses or real student data in video tools under any circumstances.

ChatGPT remains the most flexible general-purpose assistant, and it is our recommendation for student-safe deployments when configured through approved district workspace tiers. It adapts to almost any classroom task, from generating dialogue scripts to breaking down idioms. On consumer plans, however, you must never enter student personally identifiable information. You also need decent prompt skills to get consistent results across different classes, since it lacks built-in templates specifically designed for second-language acquisition frameworks.

Drawing the Line on Automated Planning

The temptation with these platforms is to let them run your entire curriculum. Resist that. AI cannot tell you that Javi spent the entire morning crying because his family is moving, or that Maria has finally figured out how to use past tense verbs when she talks about her grandmother. Those observations belong to you. Use these tools to generate the raw material, the leveled reading passages, and the first pass at a rubric. Then apply your professional judgment to make sure the lesson actually fits the human beings sitting in your room.

FAQ

Can AI do lesson planning at mixed-grade level?

It can produce a first draft and a structure. The level judgement, the standards check and the tone are teacher decisions.

Which tool is best for lesson planning?

MagicSchool AI on the evidence in this article, with Eduaide.Ai as the second opinion. Neither replaces the review step.

What is the biggest risk?

A confident, plausible output that nobody checked. The five checks below exist for that reason.

Sources

• MagicSchool AI. (2026). Feature overview and pricing tiers. Retrieved from https://www.magicschool.ai/
• Eduaide.Ai. (2026). Resource generator capabilities and plans. Retrieved from https://www.eduaide.ai/
• OpenAI. (2026). ChatGPT education safety and workspace guidelines. Retrieved from https://openai.com/
• Canva for Education. (2026). Teacher verification and classroom feature summary. Retrieved from https://www.canva.com/education/
• Synthesia. (2026). AI video generation and avatar pricing. Retrieved from https://www.synthesia.io/
• Gamma. (2026). Presentation and document generation credits. Retrieved from https://gamma.app/