Build a paraphrase drill where no content word repeats
Turns any passage into 12 paraphrase-matching items with three deliberate near-misses, and a hidden key that names each distortion.
Take this passage: [PASSAGE] Create a paraphrase-matching drill. Produce 12 items in this format: | # | Statement (paraphrased, not quoted) | Line/sentence in the passage it paraphrases | Rules: each statement must use completely different vocabulary from the passage — no content word may repeat. Three of the 12 must be near-misses that change the meaning subtly (a hedge removed, a quantity changed, a cause reversed) — mark those as TRAP in a separate answer key I should not see until I have finished. Print the drill first, then a line of dashes, then the key.
Best on ChatGPT · Claude · Gemini.
What to fill in
| Token | What to paste |
|---|---|
| [PASSAGE] | A real academic passage — a Cambridge reading text, a journal abstract, or quality long-form journalism of 700–900 words. |
What you should get back
Twelve paraphrase statements with no shared content words, then a key identifying the source sentence for each and naming the specific distortion in the three traps.
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.
| # | Statement (paraphrased, not quoted) | Line/sentence in the passage it paraphrases | | 1 | Taming equines reshaped mankind's civilization more deeply than nearly any other creature's alliance. | The domestication of the horse transformed human society more profoundly than almost any other animal partnership. | | 2 | Ancient finds from Kazakhstan's Botai group, around 3500 BCE, feature ceramic items with residues of female equine dairy and dental abrasion suggesting bridle use. | Archaeological evidence from the Botai culture of northern Kazakhstan… | | 3 | This definitely indicates the initial proof of mounted travel, even though other scholars warn that the dental marks might stem from various origins. | Some researchers argue this represents the earliest evidence of horse riding, though others caution… |
Our verdict on this run
The paraphrasing is strong — "taming equines", "female equine dairy", "dental abrasion" share no content words with the source, which is the constraint that makes this drill worth doing. But it printed the answer in the third column of the same table instead of holding it back for the key, so the drill as delivered is unusable: you can see which sentence each item maps to before you have thought about it. Ask for the statements alone first. Item 3 is also a good accidental illustration of the trap it was told to build — "definitely indicates" hardens the passage's "some researchers argue", which is exactly the overstatement that makes a statement False.
How this prompt goes wrong
It shows you the answers. The format asks for the source sentence in a third column, and models take that literally and print it beside every statement — which means the drill arrives already solved, as it did in the captured run below. Ask for the statements alone first, then the mapping. The second failure is slower: the "no content word may repeat" rule decays over a long output, so by item seven or eight the passage's own nouns start reappearing, and those items become keyword-matching rather than paraphrase recognition — the exact habit IELTS Reading punishes. Scan for lifted words before you start and regenerate the offenders. Check the three traps genuinely change meaning too, rather than only sounding different.
Tips
- Paraphrase recognition is the actual skill IELTS Reading tests, and it is the reading task AI generates most reliably.
- Scroll the key off screen before you start, or paste the drill into a separate document.
- If you can answer an item by spotting one shared word, that item is broken — regenerate it rather than counting it.
Related prompts
- ReadingGenerate True/False/Not Given items with a named trap eachForces the model to choose one of seven trap types before writing each statement, then prove the answer with the deciding sentence from the passage.
- ReadingAdjudicate one item you got wrong, and find the misreadingTakes 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.
- Writing Task 2Force the model to quote your essay before it scores itMarks your Task 2 essay in four locked steps — evidence first, descriptors second, band last — so it cannot work backwards from a flattering number.
- 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.
Common questions
Can I use a passage from a Cambridge book?
For your own private practice, yes. Do not publish or share the generated drill, since the passage remains copyrighted — this is a study tool, not a redistribution route.
Why does it keep reusing words from the passage?
Because avoiding every content word is genuinely hard, and the constraint decays over a long output. Check the later items first — that is where it slips — and ask for regenerations by number.
How is this different from real IELTS questions?
It isolates one skill. Real Reading items combine paraphrase recognition with scanning, time pressure and question-type logic. This drill strips those away so you can train the recognition itself, then put it back under timed conditions.