One common mistake ESL teachers make is spending so much class time explaining English that students have too little time to actually use it. Teachers explain grammar, introduce vocabulary, repeat instructions, model pronunciation, and correct mistakes. All of these can be valuable. The problem begins when explanation replaces practice.
Students may understand a grammar rule, recognize new vocabulary, and follow everything the teacher says — and still struggle to form a sentence when it is their turn to speak.
The goal is not to make teachers talk as little as possible. Students need clear instructions, accurate language models, and meaningful feedback. The goal is to cut unnecessary teacher talk so there is room for students to retrieve, produce, and use English themselves. This article is about the shape of a lesson that does that — and where AI can quietly help.
The Problem Is Not Teacher Talk Itself
Teacher talk is not automatically bad, and there is no single percentage that works for every ESL classroom. A beginner class may need more modeling. Young learners may need instructions repeated. A writing lesson may involve more explanation than a conversation lesson.
The more useful question is: does the teacher’s talk prepare students to use English, or does it take the place of students using English?
Sometimes the honest answer is that more teacher talk is exactly right. A class that cannot yet hear a contrast needs you to model it repeatedly. A group about to attempt something genuinely difficult needs the instructions given twice. The problem is not the minutes themselves — it is minutes spent on explanation that students were already able to infer, or that they will forget before they ever get to use it.
Consider a teacher who spends twenty minutes explaining the present perfect. The explanation may be accurate, but no one becomes confident with a tense by listening to it described. Students need to ask and answer questions, describe experiences, notice their mistakes, and try again. Knowing what a word means is not the same as retrieving it mid-conversation. Hearing a pronunciation model is not the same as producing the sound.
A Four-Stage Shape for the Lesson
Most lessons that solve this problem follow the same sequence. It is worth naming, because each stage has a job — and only one of them is the teacher talking.
1. Teacher explanation — short and load-bearing
Explain what students cannot work out from an example. Everything else can be shown rather than told: a model sentence, a demonstration, a contrast between two versions. A good test is whether you could cut the explanation in half and still have students able to start the task. If yes, cut it.
Explanation also does not have to come first, or all at once. Often it works better after students have attempted something and discovered what they are missing.
2. Preparation — move first exposure out of the performance
Much of what fills class time is first exposure: meeting the vocabulary, hearing the model, reading the text for the first time. This is the part that can happen before class or in the margins of the lesson, so class time is not spent on it.
This is where AI earns its place. A student who has already met eight target words, read a level-appropriate text, and rehearsed a question once arrives ready to use the language rather than encounter it.
3. Student performance — protect the biggest block
This is the part that gets squeezed, and it should be the largest block in the lesson. Pair and group conversation, role-plays, information-gap tasks, collaborative problem-solving — anything where students exchange real information under mild pressure.
In a sixty-minute lesson this block should rarely fall below twenty-five minutes, and it survives only if you defend it. The usual thief is the explanation running long; the second is setting up the activity so elaborately that half the block disappears into instructions. Demonstrate the task with one pair for thirty seconds instead of describing it for three minutes.
While students work, the teacher’s job changes from explaining to observing: monitoring, noting what is breaking down, and deciding what is worth addressing. That noticing is what makes the next stage possible.
4. Focused feedback — one or two things, then another go
Feedback fails when it becomes a second lecture. Pick one language point and one communication point from what you actually heard. Address those, then give students another attempt with a changed partner, audience, or complication.
The second attempt matters more than the correction. Feedback that is never used again is just more teacher talk.
Where AI Fits — and Where It Does Not
Adding an AI tool to a teacher-centred lesson does not make it student-centred. What AI can do is supply the repetition, preparation, and individual practice that one teacher cannot give thirty students inside one period. Match the tool to the stage, not the other way around:
| Stage | What AI can do | Where to read more |
|---|---|---|
| Explanation | Draft a shorter explanation, an analogy, or two contrasting examples at the right level | ChatGPT for ESL teachers |
| Preparation | Pre-teach vocabulary, supply a level-appropriate text, let students rehearse once in private | reading comprehension tools · Diffit for mixed levels |
| Performance | Extra speaking reps outside class — never a replacement for talking to a person | AI tools for speaking practice |
| Feedback | Surface-level correction on writing, individual pronunciation reps between lessons | AI pronunciation tools |
| Any stage | Nothing, if the budget is zero and the tool is not free | free AI tools with no budget |
Notice what is missing from that table: AI does not appear in the performance column as a substitute for classmates. An AI partner is patient, predictable, and always available — which makes it good rehearsal and poor practice for the unpredictability of a real conversation.
A Practical Classroom Workflow
Teachers do not need every tool in every lesson. One well-chosen tool is usually enough. A worked example:
| When | Who | What happens |
|---|---|---|
| Before class | Students | Preview eight target words · read a short level-appropriate text · rehearse three discussion questions once |
| During class | Teacher | Check vocabulary briefly · clarify only what students did not work out · put students in pairs · monitor and take notes · give focused feedback · run the task again |
| After class | Students | Repeat part of the speaking task · revise a short written response · practise two pronunciation targets |
This does not remove the teacher. It changes the role from being the source of every explanation to designing the experience, watching what students can actually do, and responding to it.
Before You Assign an AI Tool
- What specific learning need does it address?
- Does it lead to meaningful student use of English?
- Is it appropriate for my students’ age and proficiency level?
- Can all students access it — account, device, cost?
- What student information will the platform collect?
- How will I review the AI-generated content or feedback?
- What will students do with the class time this saves?
That last question matters most. Saving ten minutes of explanation does not improve a lesson unless those ten minutes become a better learning opportunity.
The Real Goal
Reducing teacher talking time is not about silencing the teacher. It is about making teacher talk purposeful and leaving enough space for students to use the language themselves.
AI can provide extra explanation, repetition, differentiated materials, and individual practice. It cannot decide what a particular group of students needs, and it cannot replace the judgment, encouragement, and responsiveness of a skilled teacher.
The most useful question is not How can AI teach this for me? It is What can students prepare or practise independently, so that our time together is spent on what only happens in a room with other people?