Make the examiner push back on every Part 3 answer
Six escalating Part 3 questions with real counter-examples and follow-ups, then a report on where your answers were too short or unsupported.
We are doing IELTS Speaking Part 3 only, on the theme of [TOPIC]. Ask me one question at a time. After each answer: - If I gave a short answer, ask "Why?" or "Can you give an example?" once. - If I made a general claim, challenge it with a counter-example. - If I hedged, press me for a position. Escalate abstraction each question: personal → societal → hypothetical → evaluative. Ask 6 questions total. Stay in examiner role. No praise, no correction, no vocabulary help during the test. After question 6, give me: the questions where my answer was under 3 sentences, the claims I failed to support, and the two questions I would score worst on — with the descriptor reason.
Best on ChatGPT · Gemini · Claude.
What to fill in
| Token | What to paste |
|---|---|
| [TOPIC] | A Part 3 theme — education, technology, the environment, work, tradition. Ideally the theme of a cue card you just practised. |
What you should get back
Six questions asked one at a time with genuine pushback between them, then a report naming your short answers, your unsupported claims, and the two weakest responses with a descriptor reason.
What it actually returned
Tested on Gemini 2.5 Flash, API, August 2026, 2026-08-06. Trimmed — this is an excerpt of a longer response, not the whole thing.
We're going to discuss the topic of education. To begin, what kind of education did you receive, and how do you think it prepared you for life?
Our verdict on this run
Short, and correct. The prompt says one question at a time, and this is what obedience looks like — it opened the theme, asked question one, and stopped, with no preamble, no praise and no offer to help. That is the behaviour you want and the behaviour most models break within two turns. The real test of this prompt is what happens at questions three and four, once you have given it a general claim to challenge; the failure mode above describes what to watch for as the exchange goes on.
How this prompt goes wrong
It breaks role and starts teaching. The most common version is a "great point!" before the next question, which destroys the pressure the prompt exists to create; the second is asking all six questions at once, which turns a conversation into a worksheet. If either happens, restate the rules and start over — mid-conversation correction rarely sticks. A subtler failure: the counter-examples get gentler as the exchange goes on, because the model is optimising for agreement. If question 6 feels easier than question 2, you got a polite chat, not sparring.
Tips
- Use voice mode if your tool has one. Typing your answers removes the time pressure that makes Part 3 hard.
- Part 3 is where Band 6.5 becomes Band 7.5, and the deciding skill is extending and defending a position — not vocabulary.
- Run it on the theme of a cue card you just answered, exactly as the real test does.
Related prompts
- SpeakingSee the same cue card answered at Band 6, 7 and 8Produces three spoken-register answers to one cue card at ascending bands, then names the concrete moves that separate each level from the next.
- SpeakingTurn a recording of yourself into countable speaking dataCounts your words per minute, filled pauses, restarts, repeated words and structure range from a transcript — and refuses to guess a band.
- Mock examinerRun a full Speaking test where the examiner never breaks roleA complete Parts 1–3 interview in examiner register — no teaching, no encouragement, no band — so it feels like the real thing.
- Mock examinerFind out what an examiner notices in your first 50 wordsTests your essay the way it is actually read — fast — separating conspicuous errors from invisible ones and naming the feature that tips the band.
Common questions
Why does it keep praising me?
Encouragement is the default behaviour and the prompt has to actively suppress it. If "no praise" is ignored, add it to your custom instructions or system prompt rather than repeating it mid-conversation — role instructions given once at the start hold better than corrections issued later.
Can I type my answers instead of speaking?
You can, and the questions still work, but you lose the part that matters. Part 3 difficulty is thinking-time pressure; typing gives you time an examiner never would.
How long should a Part 3 answer be?
Roughly three to five sentences: a position, a reason, and an example or a qualification. One-sentence answers are the most common reason a candidate stalls at Band 6.5, which is exactly what this prompt reports back to you.