From signup to band report in about an hour.
Or 15 minutes for the diagnostic. Here is exactly what happens once you press "Start" — and which AI model grades which part.
You pick a mock.
Choose a free diagnostic, a sectional drill (just speaking, just writing, etc.) or a full 2-hour TOEFL/IELTS/PTE mock. Every test in the library is real-format — no generic quizzes.
You can stop and resume any time within 24 hours. The browser locally tracks your progress so a network blip never costs you the attempt.
You take it. Real conditions.
Sections are time-boxed. Audio recording starts and stops on your behalf for speaking items. Tab-switch detection runs in the background — flagged in your report but never blocks you.
For speaking, we record one chunk per item, store it encrypted, then auto-delete after 90 days. You can hear yourself back inside the report for self-review.
AI scores each section.
Listening and reading score instantly against the answer key. Speaking gets transcribed by Whisper, scored on pronunciation by Azure, and rated by Claude against the official descriptor. Writing goes straight to Claude with the IELTS/TOEFL/PTE rubric prompt.
All four sections finish in under 5 minutes. You see scores stream in live — there is no "your report will be emailed in 2-3 days".
You read the report.
Per-skill bands, per-criterion bands on writing & speaking, and one explanatory paragraph per response. Recommended next mock based on your weakest skill.
Download a PDF for your records or share it with a tutor. The PDF carries the same per-criterion detail — nothing hidden behind a paywall.
Which AI grades which part.
We don't roll our own models. We use the best of each — and we're upfront about it.
| Skill | Item types | Provider | What it does |
|---|---|---|---|
| Listening | Write-from-dictation, Summarize spoken, Fill blanks | OpenAI Whisper | Transcribes audio; we score the transcript against the expected text. |
| Reading | MCQ single, MCQ multi, Dropdown fills | Answer key | Instant, deterministic. No AI needed. |
| Writing | Essay (Task 2), Summarize written text | Anthropic Claude | Grades against the official rubric (4 criteria for IELTS, 3 for TOEFL). |
| Speaking | Read-aloud, Repeat sentence, Describe image, Retell | Whisper + Azure | Whisper transcribes; Azure scores pronunciation; Claude rates content. |
Read the deep dive on AI-vs-human agreement for the calibration data.
Take the free 15-minute diagnostic.
Or 15 minutes for the diagnostic. Here is exactly what happens once you press "Start" — and which AI model grades which part.