IWIELTS Writing CheckerCheck my essay
Phase 0 · Calibration report

How accurate is our scoring?

Most AI writing checkers never publish how accurate they are. We do — because if you are using a tool to prepare for a high-stakes exam, you deserve to know how much trust to put in the number it gives you.

✅ Phase 0 complete · Average MAE 0.49 across 4 criteria · 28 essays

Methodology

  1. 1We collected 31 IELTS essays with official examiner-assigned band scores from British Council, IELTS Buddy, and other authoritative sources, spanning band 5.0 to 9.0.
  2. 2Each essay is scored by the model at temperature 0 with a fixed rubric derived from the official IELTS band descriptors, using few-shot anchor essays for calibration.
  3. 3We compute the mean absolute error (MAE) per criterion against the reference scores. Our target: MAE ≤ 0.5 bands on all four criteria.

What the numbers mean

  • MAE (Mean Absolute Error): the average difference between the AI score and the examiner score, in band units. Lower is better.
  • Within ±0.5 band: the share of essays where the AI estimate is within half a band of the examiner score. This is our primary quality target.
  • Pass target: all four criteria must have MAE ≤ 0.5 bands. If any criterion fails, the model is not considered calibrated.