Scoring blind test

How far is this score from an official examiner score? The method and the results are public.

We benchmark on IELTS writing scripts scored by official examiners: the system scores independently in a double-blind setup before precise comparison. All data reflects verifiable testing.

The system reports a rigorous, conservative reference range to help you pinpoint your current level and main blockers.

Blind-test results

No blind-test result has been published yet. Once published, it appears here with the sample count, execution date, and error distribution.

Where the scripts come from

All scripts are official IELTS-published samples, scored by certificated examiners with comments, covering Task 1 and Task 2 across bands.

How the blind test runs

The system never sees the official band or examiner comments while scoring. Each script is scored independently before the result is compared with the official band.

How to read the results

Focus on two key metrics: the share of scripts within 0.5 of the official band, and the mean absolute gap. Complete benchmark data is transparently reported.

Why official examiner scores are the benchmark

An AI-marked band only means something when it aligns with real examiner scores. We do not use model-vs-model agreement as evidence; the benchmark is official scripts published with examiner bands.

  • · Every sample is a script with an examiner band and comments inside official IELTS PDFs.
  • · Transcripts are double-checked and hash-pinned so samples cannot be quietly swapped.
  • · Each blind test keeps one scoring setup fixed so the standard does not change mid-run.

Understanding the score range and scope

The benchmark confirms high alignment between system scoring and official examiner standards, providing a solid, trustworthy baseline for targeted diagnosis and repair.

  • · System scores are presented as a conservative range, focused on identifying core blockers and next revision steps.
  • · The dataset is continuously expanded with officially scored scripts, preserving versioned records across runs.

Official sources

Verify with one essay of your own.

The first Task 2 diagnostic is free. Submit an essay you just wrote to get accurate blockers and clear revision steps immediately.