How Firelly Content Is Built

Reading passages & academic Listening sections

Every Reading passage, and Listening Sections 3 & 4 (the academic ones), start from a real Wikipedia article — fetched live for the topic you pick, or a random IELTS-appropriate topic if you don't. The AI paraphrases and adapts those facts into original IELTS-style prose; it never copies sentences verbatim. Easy-difficulty passages pull from Simple English Wikipedia specifically, since its plainer prose is a better starting point for Band 5-6 level than simplifying down from regular Wikipedia's more academic register.

Every grounded passage or script shows a visible attribution link to the exact source article it came from — you can always check it yourself. Listening Sections 1 & 2 (everyday conversations like booking a hotel room) are original AI scenarios by design — grounding a casual phone call in an encyclopedia article would be an odd fit.

Writing & Speaking scoring

All Writing and Speaking feedback is scored against the real, official IELTS band descriptors — the same four Writing criteria (Task Achievement, Coherence & Cohesion, Lexical Resource, Grammatical Range & Accuracy) and Speaking criteria examiners actually use, at the real 0-9 scale in 0.5 steps. These aren't invented rubrics.

Test format, timing & band conversion

Section counts, question types, and timing all match the real exam: Reading (3 passages, 40 questions, 60 minutes), Listening (4 sections, 40 questions, single-play audio), Writing (Task 1 + Task 2 with real word-count minimums), and Speaking (Parts 1/2/3 with the real 1-minute prep + 2-minute cue card timing). Raw-score-to-band conversion for Reading and Listening uses the commonly published official conversion tables, not an invented curve.

Vocabulary

The Vocabulary Builder includes full coverage of Sublist 1 (the 60 most frequent word families) of the Academic Word List (Coxhead, 2000) — a peer-reviewed, published list of 570 word families most frequent across academic texts, and the same list most legitimate academic English and IELTS/TOEFL vocabulary programs are built on. Remaining sublists and the harder literary vocabulary alongside them are a work in progress.

SAT question bank & score prediction

Every SAT practice question — in Practice, the Question Bank, Diagnostic, and Mock Test — comes directly from the College Board's own publicly accessible Digital SAT Question Bank — the same bank College Board makes available to students and educators. Nothing is AI-generated or paraphrased; stems, answer choices, and official rationales are used as published.

Score predictions are a linear estimate from your practice accuracy on the real 200–800 per-section scale, not College Board's own (unpublished) equating algorithm — treat them as a directional signal, not an official score. Mock test scoring uses the same public 200–800-per-section, 400–1600 total scale as the real Digital SAT.

What this is not

Firelly does not reproduce actual retired Cambridge, British Council, or IDP exam papers — that content is copyrighted and we don't have rights to it. Every IELTS passage, prompt, and cue card here is either grounded in genuinely open sources (as above) or AI-generated to match the real exam's format and difficulty, the same way any other prep resource writes its own practice material. Firelly is not affiliated with or endorsed by the British Council, IDP, Cambridge Assessment English, or the College Board.

One honest limitation: AI-generated items haven't been through the statistical difficulty calibration real Cambridge exam items go through before publication. Treat your estimated band scores here as a strong, directional signal for daily practice — timing discipline, question-type recognition, vocabulary, and real-criteria feedback — and pair it with official practice tests before the real exam for final calibration.