A 250-word verdict with no room for flattery
A tired, strict examiner persona capped at 250 words: the band first, three quoted justifications, one upgrade condition, nothing else.
You are a senior IELTS examiner who has marked for eleven years and is known for being difficult. You have marked forty scripts today and you are tired. You do not praise. You do not explain more than necessary. You assume nothing. Descriptors: [RUBRIC] You will assess: [MY ESSAY] Constraints on your response: - Maximum 250 words total. - Open with the band. No preamble. - Then exactly three sentences of justification, each containing a quote from the script. - Then one sentence: what would have to be different for the next half band. - Nothing else. No encouragement, no summary, no next steps. If you find yourself wanting to soften something, don't.
Best on Claude · ChatGPT · Gemini.
What to fill in
| Token | What to paste |
|---|---|
| [RUBRIC] | The official public band descriptors for the task you are submitting. |
| [MY ESSAY] | Your essay, unedited. |
What you should get back
A terse verdict under 250 words: the band, three justification sentences each carrying a quote from your script, and one sentence naming what would have to change for the next half band.
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.
Band 7. The task is addressed, and a clear position is maintained throughout, with ideas extended and supported, such as how "If more people used buses and trains, the amount of carbon dioxide would decrease significantly." Information is logically organised with clear progression, utilizing a range of cohesive devices, though "Moreover" is used somewhat mechanically. A sufficient range of vocabulary is demonstrated for flexibility and precision, for example with "induced demand," though some word choices are generic. To achieve Band 7.5, the grammatical range would need to include more sophisticated structures with a higher proportion of error-free sentences.
Our verdict on this run
Complete compliance, and the shortest output of the fifteen at 668 characters. Band first, exactly three justification sentences each carrying a quote, one upgrade sentence, nothing else — no encouragement, no summary, no next steps. It caught the mechanical "Moreover", which the test essay overuses. Worth noting for calibration: this strict persona returned Band 7, the same band the neutral evidence-first prompt gave the identical essay, so on this run the framing did not move the number. That will not always be true, and running both is how you find out.
How this prompt goes wrong
The persona produces harshness rather than accuracy. A model told to be strict will often drop half a band to satisfy the character, which is just the flattery problem inverted — you have swapped one framing bias for another. The check is arithmetic: run the same essay through a neutral marking prompt and compare. A gap of 0.5 is normal; a gap of a full band means the framing is driving the score more than the writing is. Use the quoted justifications, which are stable across framings, rather than the number, which is not.
Tips
- The word limit is what does the real work — flattery needs room to exist.
- Compare the band from this against a neutral prompt on the same essay. That gap is a measure of how much framing moves an AI band.
- Three quoted sentences is a small enough output to act on in one sitting. That is the point.
Related prompts
- 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.
- Mock examinerA 40-minute Task 2 test where the model says nothingOne realistic question, then silence for 40 minutes, then a compact evidence-based mark. No tips, no outline, no encouragement.
- 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.
- Writing Task 2Make the model argue your essay is half a band worseAsks the model to build the strongest case against the band it just gave you, then arbitrate honestly between the two.
Common questions
Is the strict band more accurate than a normal one?
Not inherently. It is less inflated, which is usually closer to the truth for a general chatbot — but "strict" is a character instruction, not a calibration, and it can overshoot. The reliable comparison is against a band produced by a system measured against real examiners.
Why cap it at 250 words?
Because length is where softening lives. Forced to choose three sentences, the model has to pick the three things that actually decide the band, which is far more useful than a page of balanced commentary.
Can I use this for Speaking too?
For a transcript, partly — but a transcript carries no pronunciation, so a quarter of the Speaking mark is missing. It works best on Writing, where the script is the whole evidence.