AI IELTS feedback
ChatGPT can’t give.
Per-criterion band estimates against the IELTS scoring criteria. Annotation-level feedback on the exact mistake, sentence, and pattern. Memory across sessions, so the same slip doesn’t keep costing you marks.
How AI grades IELTS Writing
Task 1 and Task 2 essays are graded against the four IELTS Writing band descriptors. Each criterion gets its own per-rubric estimate and a paragraph explaining where the band slipped — not a rewrite, a diagnosis. The model is calibrated against held-out candidate responses with marker bands; the headline figure is a practice band, the value is the breakdown beneath it.
- TA
Task Achievement
Whether you addressed every part of the task, with a clear position and supporting ideas — not whether your essay sounds nice.
- CC
Coherence and Cohesion
How your paragraphs are organised and how the ideas join up. Linking words on their own don't earn marks; logical flow does.
- LR
Lexical Resource
Range and accuracy of vocabulary — including collocation and word choice — judged against the band descriptors, not impressive-sounding word counts.
- GRA
Grammatical Range and Accuracy
Variety of sentence structures plus error frequency. The model flags the recurring grammar patterns pulling your score down.
How AI grades IELTS Speaking
All three Speaking parts run in one timed session — short questions, the cue-card monologue, and the discussion. The model transcribes your audio, grades against four criteria, and pinpoints the moments where hesitation, vocabulary range, or grammar accuracy pulled the band down. Pronunciation is judged against the IELTS descriptors, not a native-speaker bar.
- FC
Fluency and Coherence
Length and flow of your responses, including hesitation and self-correction. The transcript pinpoints the moments where the band slips.
- LR
Lexical Resource
Range and precision of vocabulary across the three Speaking parts. Topic-shift language and paraphrase carry weight.
- GRA
Grammatical Range and Accuracy
Variety of structures under spoken-time pressure, plus accuracy. The analysis flags grammar patterns that recur across cue cards.
- P
Pronunciation
Individual sounds, stress, intonation, and intelligibility — graded against the descriptors, not against a native-speaker benchmark.
Free diagnostic analysis for Reading and Listening
For Reading and Listening you don’t need AI grading — the answers are objective. The diagnostic value is in linking each missed answer back to the evidence and breaking your score down by question type. That’s free, full-length, and unlimited.
Transcript-linked Listening review →
Multi-accent recordings across the four canonical IELTS Listening parts. Every missed answer is linked to the line in the transcript where the cue appeared.
ReadingEvidence-highlighted Reading review →
Three Academic passages per set. After each set the analysis links every wrong answer to the sentence in the passage that proved the right one.
AI IELTS feedback,
honestly explained.
Question not listed here?
Drop us a message. We're happy to help.
After every Writing or Speaking attempt the model returns a per-criterion band estimate (four criteria each), a short paragraph explaining what cost marks against each criterion, and concrete moments in your answer that tripped the descriptor. For Listening and Reading, the analysis links each missed answer to the evidence in the audio or passage and breaks your score down by question type.
All four — Listening, Reading, Writing, and Speaking. Listening and Reading practice plus full diagnostic analysis are free for everyone. Writing and Speaking AI evaluations are on the Pro plan because each evaluation runs against multiple model passes.
Each estimate is a practice band, not an official score. The model is calibrated against the public IELTS band descriptors and a held-out set of marked candidate responses. The diagnostic value is the per-criterion breakdown beneath the band — what cost the marks — rather than the headline number.
No. ChatGPT rewrites essays. Zoju grades them — against the IELTS scoring criteria, with per-criterion feedback on what cost the marks, and memory across sessions so the same slip doesn't keep costing you. Different tools, different jobs.
Because you can't fix what you can't name. ChatGPT will rewrite your essay into something that scores better — for the rewrite. Your next essay will have the same problems, because nothing learned across the conversation, and the rewrite didn't tell you which IELTS criterion you missed or why. Zoju's job is the diagnosis: the recurring slip, the criterion it lives in, the moment in your answer where it cost the marks. The band lift is yours, not the rewrite's.
No. Free Listening and Reading practice — full sets, full analysis, unlimited — is available without a card. The Pro plan adds 20 Writing and 20 Speaking AI evaluations per month at $15/month, with the LAUNCH20 promo bringing the first three months to $12/month.
You can read every public page without an account. Starting a saved practice session — so the analysis can be stored and your patterns tracked across sessions — requires a free account. The Listening and Reading flows do not require payment.
Practice. Diagnose. Repeat.
Free Listening and Reading practice with full analysis. Pro adds AI evaluations for Writing and Speaking.
Start free →No card required. Free forever.
Also free: see Zoju pricing