These are IELTS prompts for ChatGPT, Claude and Gemini that load the official band descriptors, force the model to quote your work before it judges it, and fix the output format so two runs are comparable. Every prompt shows the real output it produced and the specific way it goes wrong.
Almost every “IELTS ChatGPT prompt” online is one sentence: act as an IELTS examiner and score my essay. That produces a number with no relationship to the descriptors, and a different number every time you run it. Four things fix that, and every prompt in this library uses them.
Models paraphrase the band descriptors from memory and the paraphrase drifts. The official public descriptors are a free download from ielts.org, and pasting them is the difference between descriptor-anchored marking and a guess.
A model made to quote your text before assigning a band cannot work backwards from a number it already picked. Every marking prompt here enforces that order.
A fixed table or field list makes two runs comparable. Without one, the same essay produces differently-shaped feedback each time and you cannot tell whether you improved.
Praise earlier in a conversation pulls the next band up. This is the single most common reason an AI band is too generous.
Marking prompts that quote your essay before they score it
Marks your Task 2 essay in four locked steps — evidence first, descriptors second, band last — so it cannot work backwards from a flattering number.
Best on Claude · ChatGPT · Bands 6.0–8.0
Marks a reference essay whose real band you already know, forces a recalibration, then marks yours against that same standard.
Best on ChatGPT · Claude · Bands 6.0–7.5
Asks the model to build the strongest case against the band it just gave you, then arbitrate honestly between the two.
Best on Claude · ChatGPT · Bands 6.0–7.5
Maps your controlling ideas, referencing chains and linker repeats — the criterion most learners cannot self-assess — with grammar and vocabulary explicitly off-limits.
Best on Claude · ChatGPT · Bands 6.0–7.5
Runs a yes/no scorecard over the two most formulaic paragraphs in Task 2, quotes the evidence, then rewrites both in your own voice.
Best on ChatGPT · Claude · Bands 5.5–7.0
Prompts that measure what a transcript can actually show
Counts your words per minute, filled pauses, restarts, repeated words and structure range from a transcript — and refuses to guess a band.
Best on ChatGPT · Claude · Bands 6.0–7.5
Produces three spoken-register answers to one cue card at ascending bands, then names the concrete moves that separate each level from the next.
Best on Claude · ChatGPT · Bands 6.0–8.0
Six escalating Part 3 questions with real counter-examples and follow-ups, then a report on where your answers were too short or unsupported.
Best on ChatGPT · Gemini · Bands 6.5–8.0
Drill generators, and an adjudicator for the answers you got wrong
Turns any passage into 12 paraphrase-matching items with three deliberate near-misses, and a hidden key that names each distortion.
Best on ChatGPT · Claude · Bands 5.5–7.0
Forces the model to choose one of seven trap types before writing each statement, then prove the answer with the deciding sentence from the passage.
Best on Claude · ChatGPT · Bands 6.0–7.5
Takes a single wrong answer and returns the deciding sentence, the False-versus-Not-Given logic, the exact feature you misread, and a 10-second in-exam test.
Best on Claude · ChatGPT · Bands 6.0–7.5
Personas that stay in role, stay strict, and stay quiet
A complete Parts 1–3 interview in examiner register — no teaching, no encouragement, no band — so it feels like the real thing.
Best on ChatGPT · Gemini · Bands 6.0–8.0
A tired, strict examiner persona capped at 250 words: the band first, three quoted justifications, one upgrade condition, nothing else.
Best on Claude · ChatGPT · Bands 6.0–7.5
One realistic question, then silence for 40 minutes, then a compact evidence-based mark. No tips, no outline, no encouragement.
Best on ChatGPT · Claude · Bands 5.5–7.5
Tests your essay the way it is actually read — fast — separating conspicuous errors from invisible ones and naming the feature that tips the band.
Best on Claude · ChatGPT · Bands 6.0–8.0
There is no single best one, because the jobs are different: marking an essay, drilling Reading, or running a Speaking mock each need a different structure. What every prompt worth using has in common is four things — the official descriptors pasted into context rather than recalled, a fixed output format, evidence demanded before any judgement, and a calibration reference where one exists. A one-sentence "act as an IELTS examiner" prompt has none of those and returns a different band every time you run it.
It can produce a plausible band, which is not the same as an accurate one. The number moves with how the prompt is framed — the same essay can come back 6.5 from a strict persona and 7.5 from a neutral one. In our own study of 1,200 essays, purpose-built AI grading matched the human examiner consensus within ±0.5 bands 94.2% of the time; a general chatbot with a pasted rubric is looser than that. Use these prompts for the reasoning and the quoted evidence, not for the digit.
Claude and ChatGPT both handle a full rubric plus an essay reliably, which matters for marking prompts. ChatGPT and Gemini have the better voice modes for Speaking practice. Gemini is the one to use when a chart needs reading for Academic Task 1. In practice the tool matters far less than whether your prompt loads the descriptors and forces evidence before judgement.
Yes, all of them, on any assistant. Several are long enough that a free-tier context limit can truncate them silently, which usually shows up as the model inventing descriptor wording — each page notes when that is a risk.
Because a prompt library that only shows successes is useless the first time a model behaves differently for you. Every page here carries a real, dated run and a "how it goes wrong" section, so you can tell the difference between the prompt failing and your essay being weak.