Methodology

How the marking works

Most free checkers tell you what you want to hear. This page explains — precisely, with nothing hidden — how Band 7 Lab arrives at your band, so you can decide whether to trust it.

1. Your essay is marked against the official public band descriptors

When you submit, your full essay is assessed against the public IELTS Writing band descriptors — the same criteria real examiners use: Task Response (Task Achievement for Task 1), Coherence & Cohesion, Lexical Resource, and Grammatical Range & Accuracy. The complete descriptor scales, band by band, are supplied to our AI examiner on every single marking — it doesn't score from general impressions.

2. Four separate scores, each with evidence from your own writing

Each criterion gets a whole-band score chosen by matching your script against the descriptor wording. For every criterion, the examiner must quote specific wording from your essay as evidence — you'll see exactly which sentences earned or cost you the band. Your overall band is the average of the four, rounded to the nearest half band, the same arithmetic the real test uses.

3. The length penalty is enforced in code, not left to the AI's mood

An under-length essay cannot fully answer the task, so the real exam penalises it — and so do we, deterministically:

• Task 2 under 250 words (Task 1 under 150): Task Response is capped at 6, no matter how good the writing is
• Severely short — Task 2 under 200 (Task 1 under 120): capped at 5

This cap is applied by our software after marking, so it can never be "charmed" out of the score. We added it after testing showed that AI examiners, like human ones, are tempted to forgive length when the English is good.

4. Calibrated to never inflate — deliberately

The examiner runs under strict calibration rules: Band 7 grammar requires genuinely frequent error-free sentences; Band 7 vocabulary requires less-common words used with real control; a good essay is a 7, not an 8; and only the four official criteria are ever assessed — punctuation style, formatting and spacing are never penalised, because IELTS doesn't assess them.

5. It also remembers you

With a free account, the examiner is shown your recurring weaknesses from previous essays on every new marking — and must state whether you've repeated each habit (with proof quoted from the new essay) or beaten it. Your misspellings feed a personal spaced-repetition vault. This is how the tool coaches rather than just scores.

What this is — and isn't

• This is an AI examiner (built on Claude, a frontier language model) marking under the constraints above — not a human examiner, and not an official IELTS score.
• Band scores here are practice estimates for study. Only the IELTS partners issue real results.
• In our internal regression testing the same essay receives the same band on repeated marking, and known-weak essays score low — we test this every time we change the examiner. We publish no accuracy percentage because we haven't yet run a public benchmark worth citing; when we do, it will be on this page.
• Band 7 Lab is independent — not affiliated with, or endorsed by, the British Council, IDP Education, or Cambridge. "IELTS" is their registered trademark, used here only to describe the exam.