How VoyaLang scores your IELTS writing.
VoyaLang grades IELTS writing against the four official assessment criteria and reports a band for each one, plus an overall band and the specific sentences behind each judgment. This page explains how that scoring was built and what it can and cannot tell you.
The four criteria are the whole score.
IELTS Writing is not marked as one impression. It is marked on four separate criteria, each carrying equal weight.
- Task Achievement, or Task Response Did you answer the question that was actually asked, cover every part of it, and support what you claimed?
- Coherence and Cohesion Is the writing organized, does it paragraph sensibly, and does it connect ideas without leaning on a list of linking words?
- Lexical Resource Is the vocabulary precise and appropriate, rather than merely uncommon?
- Grammatical Range and Accuracy Is there a genuine range of structures, and are they controlled?
VoyaLang reports all four separately, because an overall band on its own does not tell you what to fix. Two candidates can sit at the same overall band for completely different reasons and need opposite advice.
How the scoring was built.
We did not start from a theory of how IELTS should be marked and ship it.
We started from real writing that had already been graded against the official criteria, and worked backwards. For every task type we ran the scoring against that graded writing, compared what our system produced with the grade the writing already carried, changed one thing, and ran it again. Then again. The version that shipped is the one whose grades landed closest to the grades those writings already had.
That process was run separately for each task type, because they are not the same job. Describing a process diagram is not the same skill as arguing a position, and an approach tuned on essays does not transfer to charts. Academic Task 1 goes further still: a line graph, a pie chart, a map and a process diagram are each assessed on their own terms, because what counts as covering the task is different for each one.
The system is designed, built and tuned by an AI engineer rather than assembled from a generic template, and it is monitored so that it keeps behaving the way it was measured to behave.
We keep the specific rules to ourselves. What we will tell you is the shape of the work: iterate against writing that already has a grade, keep what closes the gap, discard what does not.
What we do not claim.
- We do not replace a certified examiner.A human examiner may reach a different conclusion on the same piece of writing. Examiners themselves do not always agree with each other.
- We do not guarantee your exam band.VoyaLang tells you where your writing sits today and what is holding it back. It is a practice instrument, not a prediction and not a verdict.
- We do not publish an accuracy percentage.We have not found a way to produce one that would be honest rather than flattering, so we describe the method instead and let you judge the output.
If you disagree with a score.
Tell us. There is a feedback control on every result. We read them, and disagreements on specific pieces of writing are the single most useful thing we receive, because they are exactly the input the tuning above runs on.
VoyaLang is not affiliated with or endorsed by IELTS, the British Council, IDP, or Cambridge Assessment English.