ChatGPT prompts for ESL speaking work best when they give the model a clear role, a level, and a turn-taking structure instead of a vague request like “give me a speaking activity.” Below are 20 ready-to-use prompts organized by CEFR level and activity type, each with a one-line note on why it actually gets students talking instead of producing a wall of text to read aloud.
Most teachers’ first attempts at using ChatGPT for speaking practice fail for a predictable reason: they ask for “a conversation” and get a script, or ask for “a role play” and get stage directions instead of something a student can actually perform. The prompts here are written to avoid that trap, and they are grouped so you can grab one in the thirty seconds before class starts. We will also cover the two class formats (1:1 and group), the most common pitfalls, and how these prompts fit alongside a proper lesson plan.
Beginner speaking prompts need heavy scaffolding: short turns, concrete topics, and a model sentence structure baked into the prompt itself, or the output will overshoot what an A1 or A2 student can actually say out loud.
“Write 8 simple ‘this or that’ questions for A1 English learners using only present simple tense and vocabulary from the topics food, animals, and free time. Format each as a short question a teacher can read aloud.”
Why it works: constraining the grammar and topic list prevents ChatGPT from generating anything above an A1 reading level, and the “read aloud” instruction keeps answers short.
“Create a 10-line picture description warm-up for A1 students describing a busy street scene, using only ‘there is / there are’ and basic adjectives.”
Why it works: picture description with a locked grammar structure gives every student the same sentence frame, which lowers anxiety for true beginners.
“Write 6 two-line role play scripts for A2 students ordering food at a restaurant, with blanks for the student to fill in their own choice.”
Why it works: fill-in-the-blank role plays give structure while still requiring a real spoken choice, which is the sweet spot for A2 confidence-building.
“Generate 8 A2-level ‘would you rather’ questions about daily routines that a pair of students can ask each other, each with two clearly contrasting options.”
Why it works: forced-choice questions are almost impossible to answer with silence, which is exactly what you want from a warm-up.
“Write a 12-line A2 dialogue between a shop assistant and a customer buying clothes, then remove the customer’s lines and replace them with prompts in brackets for the student to improvise.”
Why it works: half a real script plus improvised gaps trains students to produce language rather than just recite it, without the intimidation of a blank page.
Intermediate students can handle opinion and hypothetical language, so this is where debates, structured discussions, and light role plays with a twist start to work well.
“Write 6 B1-level discussion questions about social media use that require students to give an opinion and one supporting reason, avoiding yes/no phrasing.”
Why it works: banning yes/no phrasing forces a full sentence response, which is the single biggest speaking-practice upgrade for B1 classes.
“Create a 4-round mini debate structure for B2 students on the topic ‘remote work is better than office work’, with one opening statement prompt, two rebuttal prompts, and one closing statement prompt per side.”
Why it works: a structured debate format gives quieter students a script for turn-taking while still requiring genuine argument in the moment.
“Write a B1 role play for two colleagues disagreeing about which restaurant to choose for a work lunch, including 3 useful phrases for polite disagreement at the top.”
Why it works: pre-teaching the functional language (polite disagreement) inside the prompt means the role play also becomes a mini language lesson.
“Generate 8 B2-level hypothetical questions using second conditional structures about lifestyle changes (e.g. ‘If you could live in any city, where would you choose and why?’).”
Why it works: hypothetical questions naturally produce longer, more complex responses than factual questions, which is exactly the stretch a B2 class needs.
“Write a picture-based storytelling prompt for B1 students: describe an ambiguous photo of a busy scene and ask 3 follow-up questions about what might have happened before and after.”
Why it works: ambiguity is the key ingredient, since a scene with an obvious single interpretation kills discussion, while an open scene invites competing stories.
“Create a B2 negotiation role play between a tenant and landlord discussing a rent increase, with each side given 3 private goals not visible to the other.”
Why it works: information gaps (each side knowing something the other does not) are one of the most reliable communicative-speaking mechanisms in ESL teaching.
Advanced students need nuance, abstraction, and pushback, so prompts at this level should ask for genuine complexity rather than simplified topics dressed up in harder vocabulary.
“Write 5 C1-level discussion questions on the ethics of AI in hiring decisions, each requiring the speaker to consider a counterargument before giving their final position.”
Why it works: requiring an acknowledged counterargument pushes C1 students past simple opinion-stating into the more sophisticated concession-and-rebuttal structures examiners and real conversation both reward.
“Create a C1 debate on ‘universal basic income should replace unemployment benefits’ with 3 strong points and 2 counterpoints per side, phrased as prompts rather than full arguments.”
Why it works: prompts rather than full arguments force the student to build the actual sentences, keeping this a speaking task instead of a reading task.
“Write a C1 role play between a journalist and a company spokesperson responding to a product recall, including 4 increasingly pointed follow-up questions.”
Why it works: escalating follow-up questions simulate the real pressure of unscripted spoken exchange, which is exactly the skill C1 speaking assessments test.
“Generate 6 abstract ‘this or that’ philosophical questions for C1 students (e.g. ‘is it better to know too much or too little about your future?’) with no factual right answer.”
Why it works: genuinely unresolvable questions keep advanced students engaged far longer than factual ones, since there is always another angle to argue.
“Write a C1 fluency drill: 8 idiomatic expressions related to work and career, each followed by a question the student must answer using that idiom naturally.”
Why it works: forcing idiom use inside a real answer, rather than in an isolated matching exercise, is what actually moves idioms from passive to active vocabulary.
These last five work across a wider level range and are useful when you need something fast, or when you are teaching a mixed-ability group.
“Write a 5-minute warm-up: 6 rapid-fire personal questions on the topic of weekend plans, ordered from easiest to hardest.”
Why it works: ordering questions from easiest to hardest lets every student succeed early, which raises participation for the harder questions that follow.
“Create a 1-minute fluency drill: a single topic sentence starter (‘The best thing about my hometown is…’) that a student must keep talking about for 60 seconds without stopping.”
Why it works: timed, unbroken speech is one of the few exercises that directly targets hesitation and filler-word habits rather than grammar accuracy.
“Generate a picture description activity using an office meeting scene, with 3 questions at A2 level and 3 follow-up questions at B2 level on the same image.”
Why it works: layering two difficulty tiers onto one image lets a single prompt serve a mixed-level class without extra prep work.
“Write 6 ‘agree or disagree’ statements about technology and daily life suitable for a mingling activity where students walk around comparing answers.”
Why it works: mingling formats multiply speaking time per student far beyond what a single whole-class discussion can achieve in the same number of minutes.
“Create a storytelling chain prompt: an opening line for a story that students must continue one sentence at a time around the group, with a twist requirement introduced halfway through.”
Why it works: chain storytelling keeps every student listening actively, since they cannot contribute their sentence without tracking what came before.
Use this table to quickly match a prompt style to the class in front of you, whether that is a single online learner or a full group.
| Activity type | Best level range | 1:1 or group | Typical prep time |
|---|---|---|---|
| Picture description | A1 to B2 | Both, works well 1:1 | 2 to 3 minutes |
| Role play with script gaps | A2 to C1 | Both, ideal for pairs | 3 to 5 minutes |
| Structured debate | B2 to C1 | Group (4+ students) | 5 to 8 minutes |
| Warm-up questions | A1 to C1 | Both | 1 to 2 minutes |
| Fluency drills | B1 to C1 | Both, strong 1:1 fit | 2 minutes |
| Mingling / agree-disagree | A2 to B2 | Group only | 3 to 4 minutes |
Data source and methodology: internal classification based on activity structure and typical ESL classroom use, cross-checked against CEFR level descriptors for spoken production and interaction as of 2026.
ChatGPT is a strong prompt generator, but it is not a classroom management tool, and a few recurring mistakes show up whenever teachers first start building speaking activities this way.
| Pitfall | Fix |
|---|---|
| Vague prompts producing a script instead of a speaking task | Always specify level, topic, grammar constraint, and output format (questions, role play, debate structure) in the same prompt. |
| Output above the class’s actual level | Name the CEFR level explicitly and restrict grammar or vocabulary in the prompt rather than editing the output afterward. |
| Students reading answers off a screen instead of speaking naturally | Ask ChatGPT for prompts and question stems, not full scripted answers, so students must produce their own language. |
| Repetitive topics across sessions | Keep a running list of topics already used and explicitly exclude them in the next prompt. |
| No natural follow-up when a student finishes early | Ask for 2 to 3 built-in follow-up questions per prompt so early finishers keep talking instead of going quiet. |
| One-size-fits-all prompts in mixed-ability classes | Request two difficulty tiers on the same topic in a single prompt, as in the mixed-level section above. |
A strong speaking prompt is one piece of a lesson, not the whole thing, and this is where teachers who stop at “copy the prompt into ChatGPT” leave a lot of value on the table. A complete speaking lesson still needs a warm-up, a short language focus (even five minutes pre-teaching two or three useful phrases), the speaking activity itself, and a feedback stage where you note recurring errors to address next class.
Rather than building that structure from scratch around each prompt, our free lesson plan generator can build the full lesson shell (warm-up, language focus, main activity, feedback stage, timing) around any of the prompts above in under a minute, so you are not manually formatting a lesson plan every time you want to run a new speaking activity. It is free to use and does not require an account.
If you already have a coursebook speaking activity that feels flat or dated, our AI materials adaptor can rewrite it to a different level or refresh the topic while keeping the same grammar target, which is often faster than writing a brand new activity from scratch.
The prompts above work in both formats, but the way you deploy them should shift. In a 1:1 online class, fluency drills, picture description, and role plays where you play one character yourself get the most speaking time per minute, since there is no group dynamic to manage and every second of class time is the student’s turn to talk. Debates and mingling activities do not translate well to 1:1 unless you are willing to argue the opposing side yourself, which works occasionally as a change of pace but should not be your default.
In a group class, the calculation flips. Structured debates, mingling activities, and chain storytelling scale beautifully because they multiply speaking opportunities across the room instead of bottlenecking through one teacher-student exchange. The main risk in group settings is uneven participation, so building in the two-tier difficulty prompts and explicit turn-taking structures described above matters more here than in 1:1 contexts, where turn-taking is automatic.
The teachers who get the most out of ChatGPT prompts for ESL speaking are not the ones with the cleverest single prompt, they are the ones who build a small, reusable prompt library organized by level and activity type, the same way this article is organized. Save the ones that work well with your specific classes, tweak the topic each time you reuse them, and retire ones that consistently fall flat with your students. Over a term, that habit saves far more prep time than searching for a new speaking activity from scratch every single class.
If you want a broader look at what else is available beyond a single prompt library, our AI tool overview rounds up the free tools we cover most often, including the lesson plan generator and materials adaptor mentioned above.
It is also worth being honest about where ChatGPT genuinely helps and where it does not. It is excellent at generating volume and variety of prompts quickly, and at tailoring language to a specified level once you give it clear constraints. It is not a substitute for reading the room: if a debate prompt is falling flat because your group is shy that day, no prompt engineering fixes that, and switching to a lower-pressure warm-up format is still the right call. Use these prompts as a fast starting point, not a script to follow rigidly regardless of how the class is actually going.
Speaking practice lives or dies on structure, not enthusiasm alone, and a well-built ChatGPT prompt gives you that structure in seconds instead of the twenty minutes it used to take to write a role play or debate format from scratch. Start with two or three prompts from the level you teach most, run them this week, and build your library from there.
The best prompts specify the CEFR level, the topic, a grammar or vocabulary constraint, and the output format (role play, debate, warm-up questions) all in one request. Vague prompts like “give me a speaking activity” tend to produce scripts to read rather than genuine speaking tasks.
Yes. ChatGPT is well suited to generating role plays, especially when you ask for scripts with gaps or improvisation points rather than a fully written dialogue, since gaps force students to produce their own language instead of reading a finished script.
Name the exact CEFR level and restrict the grammar structures or vocabulary range directly in the prompt, rather than generating open-ended content and simplifying it afterward. This keeps the output matched to what your class can realistically produce out loud.
Many are, particularly fluency drills, picture description, and role plays where the teacher plays one character. Group-oriented formats like debates and mingling activities do not translate as well to 1:1 lessons unless the teacher is willing to argue the opposing side.
Two or three is usually enough: one warm-up, one main activity, and a backup in case the group finishes early or a topic falls flat. Building a small reusable library by level and activity type saves far more time than generating a new prompt for every single class.
No. A prompt is one component of a lesson, not the whole lesson. A complete speaking lesson still needs a warm-up, a short language focus, the speaking activity itself, and a feedback stage to address recurring errors afterward.
Start with A1 or A2 prompts that use heavy scaffolding: short turns, concrete topics, and a locked grammar structure. Move to open-ended, hypothetical, or debate-style prompts only once students are comfortably producing full sentences at the lower levels.
The past simple is the English verb form used for an action or state completed at a finished time in the past. Most verbs form it by adding -ed, while a closed set of common verbs change their form instead. It is the ordinary narrative form of English and the first past form most learners meet.
The present perfect is the English verb form made with have or has plus a past participle, used to link a past action to the present moment. It covers experience up to now, a situation that began in the past and still continues, and a past event whose result matters now. It never takes a finished time reference.
