If you're an ESL teacher in 2026, you've probably used AI to plan a lesson. Maybe you asked ChatGPT for discussion questions, had Claude generate a debate topic, or used Perplexity to find a reading text. AI has made lesson planning faster, easier, and more creative.
But here's the thing nobody talks about: AI is excellent at generating speaking content and terrible at managing speaking practice. The gap between "here are some great questions" and "every student in this room is now practising in pairs with appropriate questions, timers, and partner rotation" is enormous — and that gap is where the real teaching happens.
What AI does brilliantly
Question generation. Need 20 discussion questions about technology for B2 students? AI produces them in seconds, often better than what you'd find in a textbook. The quality is consistently high, the topics are current, and you can specify exactly the level and format you need. Lesson planning. AI can structure a full speaking lesson in minutes — warm-up, main activity, extension, debrief — with appropriate scaffolding for your target level. For new teachers especially, this is transformative. Differentiation. Need the same topic at three different CEFR levels? AI adapts the same core question into A2, B1, and B2 versions instantly. That used to take 20 minutes of manual rewriting. Content variety. AI never runs out of ideas. When you've recycled the same discussion topics for the third term running, AI generates fresh ones that feel relevant and current.What AI can't do in the classroom
Here's where the limitations become practical problems:
AI can't pair your students. It can suggest pairing strategies, but it can't look at your class of 28, apply random or strategic pairing, respect conflict rules, and display the result on screen for everyone to see. AI can't time the activity. It can tell you "give students 2 minutes" but it can't run a visible countdown that the whole room works to, creating shared rhythm and natural transitions. AI can't track question history. If you use AI to generate questions each lesson, you'll eventually repeat topics without realising it. After a term, students have answered "Do you prefer working from home?" three times — and they remember. AI can't manage simultaneous pair work. The real magic of speaking practice is 15 pairs talking at the same time. AI generates the questions; you still have to display them, pair the students, time the rounds, and rotate partners manually. AI can't replace a real listener. Chatbots can hold a conversation, but they can't replicate what happens when a human partner doesn't understand you and you have to rephrase. That negotiation of meaning — the confused look, the "what do you mean?", the reformulation — is where real acquisition happens.Tool tip: YapYapGo was designed to bridge exactly this gap. It combines AI-generated questions with classroom management features — automatic pair shuffling (random, stretch, matched, or mixed), built-in countdown timers, cross-session question tracking, and six distinct speaking modes. The AI generates the content; the tool manages the classroom practice. Free to start.
The real power: AI + structured classroom practice
The most effective approach isn't AI or traditional classroom speaking practice. It's both — used for what each does best.
Use AI for content creation: Question generation, topic adaptation, level differentiation, lesson planning. This saves you time and gives you better raw materials. Use a classroom tool for delivery: Pairing, timing, rotation, question display, and history tracking. This turns those raw materials into structured, equitable speaking practice where every student participates.Think of it like cooking: AI is the recipe book, always full of ideas. But you still need a kitchen — the physical space with tools and structure — to turn ingredients into a meal.
What the research supports
The evidence on speaking development consistently points to the same factors: high-volume practice, varied partners, time pressure, and genuine communicative purpose. AI chatbots can provide conversation practice for individual students — and that's genuinely useful for homework and self-study.
But classroom pair work provides something AI can't: peer interaction with real communication breakdowns. When your partner doesn't understand you, you have to adapt in real time. When your partner says something unexpected, you have to respond spontaneously. When you're paired with someone new, you encounter different vocabulary, different ideas, and different communication styles.
Research on peer interaction consistently shows that it creates more speaking opportunities, more negotiation of meaning, and more language-related episodes than teacher-fronted or AI-mediated formats. The classroom isn't obsolete — it's the one environment where real-time, multi-partner communication practice happens at scale.
The teacher's role hasn't changed
With all the AI tools available, the teacher's core job during speaking practice is exactly what it's always been: circulate, listen, note patterns, and give feedback after the activity.
The tools have gotten better at handling the logistics — generating questions, pairing students, timing activities. That's a good thing, because it frees you up to do the thing no tool can do: use your professional judgment to diagnose common errors, notice which students need encouragement, and decide what to address in the debrief.
YapYapGo handles the logistics so you can focus on the teaching. Six speaking modes, automatic pairing, built-in timers, AI-generated questions, and a question bank that tracks history across sessions. It's free to start — and it turns every lesson into a structured speaking practice session without any prep.Sources:
- Sato, M. & Ballinger, S. (Eds.) (2016). Peer Interaction and Second Language Learning. John Benjamins. — Peer interaction creates more speaking opportunities than AI-mediated or teacher-fronted formats.
- Long, M. (1996). The Role of the Linguistic Environment in Second Language Acquisition. Handbook of Second Language Acquisition. — The Interaction Hypothesis: negotiation of meaning drives acquisition.
- Sato, M. & Lyster, R. (2012). Peer Interaction and Corrective Feedback. Studies in Second Language Acquisition. — Peer practice increases fluency and accuracy.
- Smith, M., Vickrey, T., Pass, L., & Baird, R. (2012). Combining Peer Discussion with Instructor Explanation. CBE—Life Sciences Education. — Peer + teacher is more effective than either alone.