IELTS Reading AI prompts

AI is unreliable at generating IELTS Reading questions and reliable at explaining them. That asymmetry shapes this category. Ask a model for ten True/False/Not Given items and a predictable share will be invalid — usually a statement labelled Not Given that the passage actually contradicts, which is False. Ask the same model why a specific item you got wrong is Not Given rather than False, with the passage in front of it, and the explanation is usually excellent.

So the prompts here do two different jobs. The generators build drills with structural constraints that make their output auditable: a paraphrase drill where no content word may repeat, and a True/False/Not Given generator that has to name one of seven trap types before writing each statement and then quote the sentence that decides the answer. That quoted sentence is the audit trail — if the key cannot point at a sentence, the item is broken, and you should trust your own reading over the key.

The adjudicator does the other job. You feed it one item you got wrong, the surrounding paragraphs and both answers, and it returns the deciding quote, the False-versus-Not-Given logic, the specific feature you misread — a hedge, a scope word, a comparison never made, or your own world knowledge — and a ten-second test you could have run in the exam. That last part is the one that transfers.

Underneath all of it is a single skill: paraphrase recognition. IELTS Reading is not a vocabulary test and it is not a speed-reading test; it is a test of whether you can tell that two differently-worded sentences mean the same thing, and whether you can tell when they nearly do but do not. Every trap type in the taxonomy is a variation on that. The traps our own practice engine names on your wrong answers — Partial Truth, Extreme Language, Outside Text, Opposite Meaning — are the same family.