Responsible use
Writeiq supports teachers to assess writing. It does not mark, grade, or decide a student's achievement. A teacher reviews the evidence and confirms the result, and that teacher-confirmed judgement is the official one. This page sets out how Writeiq is meant to be used, what it does with your data, and the concerns we take seriously.
Writeiq is a teacher judgement support system. It applies the Integrated Writing Assessment Framework (IWAF) to a piece of writing and returns a criterion-by-criterion analysis, draft feedback, and a suggested Gradual Release of Responsibility (GRR) lesson. A teacher reads that evidence, applies their professional judgement, and confirms the final moderated score and the next teaching steps. The teacher-confirmed score is the official assessment judgement. Anything Writeiq produces before that point is a draft for the teacher to review.
Writeiq does not mark, grade, or determine a student's achievement, and it does not write the report a family reads. A teacher does. We have built the product so this holds in the software, not only on this page: the original analysis is kept as evidence and cannot be quietly overwritten, and the score that counts is recorded as a decision the teacher made.
Analyses a piece of writing against the IWAF criteria. Drafts feedback in plain language. Suggests a GRR lesson aimed at the highest-priority criterion. Surfaces patterns across a class or cohort for the teacher and school to interpret.
Reviews the analysis as evidence. Adjusts any criterion score, with a reason recorded. Confirms the final moderated score. Decides what feedback the student and family actually receive, and what to teach next.
Produce a final score on its own. Decide achievement. Write for the student. Send a student or family a result that a teacher has not confirmed.
There is always a teacher between the analysis and the result.
Our database and stored school data are hosted in our Supabase project, deployed in the Sydney region (ap-southeast-2). Supabase provides encryption in transit and at rest; its security and regional documentation is linked in Sources below. More detail is on our Security and Privacy page.
Writeiq uses a large language model (the Claude model family, provided by Anthropic) to produce its analysis. Edsthetic does not use student writing to train any model. Anthropic's published position is that data sent through its commercial API is not used to train its models; that statement is linked in Sources below.
For typed writing, no student names, class names, or school identifiers are sent to the model provider. The optional Vision feature, used to read handwritten work, sends the page image plus whatever the student wrote at the top to match it to the right student. A school can turn Vision off.
Schools remain the holder of their own rubric source documents. Edsthetic stores a structured configuration of how a school's rubric maps to the framework, not a long-term copy of the school's source files.
Writeiq's analysis is produced by a statistical model. It is useful, and it is not infallible. We are deliberately honest about this, because a tool that claimed to be always right would invite teachers to stop checking, and that is exactly what we do not want.
The same piece may not receive an identical analysis every time. This is one reason a teacher confirms the result rather than accepting it automatically.
We calibrate agreement with teacher judgement to sit within the range two experienced teachers reach with each other, not above it. Claiming better-than-human accuracy would be misleading.
Where the analysis systematically diverges from teachers' confirmed scores on a criterion, that is visible to us and can trigger a review or a rollback of the framework version.
Independent research on this kind of software raises real concerns. We do not think they disappear because a teacher is in the loop. The teacher's role is central to how we respond to each.
Automated scoring can push writing toward what a rubric rewards and flatten a student's voice. Writeiq keeps the teacher in charge of the judgement, treats the analysis as evidence rather than a verdict, and is criterion-referenced rather than style-prescriptive. The risk is real, and the teacher is the safeguard.
The underlying model is trained on a corpus we do not control and cannot fully audit. We respond with teacher review of every result, calibration against the Australian curriculum, an external standard-setting study, and drift monitoring. We do not claim the model is free of bias.
Any tool like this risks teachers or students leaning on it and losing practice. Writeiq is designed to surface evidence for a teacher to reason about, not a verdict to accept, and we ask schools to treat it as support, not a replacement for professional judgement.
Writeiq's analysis is currently focused on writing in English and does not support First Nations languages or dialects. This is a genuine limitation with equity implications, and we name it rather than work around it.
The evidence is that students need more time writing, and that peer review matters. Writeiq works on writing students have already done. It should not reduce writing time, and its moderation features support teachers rather than substitute for students reviewing each other's work.
Where a school uses Writeiq with students, it does so on an opt-in basis with a clear collection notice, not by default.
Declining must not disadvantage a student. A school can assess that student's writing through ordinary teacher marking, without Writeiq, and without the student being singled out.
Whether and how Writeiq is used is the school's and its community's decision. Our job is to be plain about what it does so that decision is informed.
Because the official score is the one a teacher confirmed, questioning a result means engaging a human judgement, through your school's normal process. There is no separate machine verdict to dispute.
An honest statement names what is outstanding. Writeiq is working toward participation in Safer Technologies 4 Schools (ST4S), with a readiness check planned and a badge targeted for 2027. We have not yet completed an independent, third-party audit of the analysis for fairness. Teacher review, calibration against the Australian curriculum, and drift monitoring are what stand in the meantime, and we say so plainly rather than implying the work is done.
We wrote this position against public policy, technical documentation and the risks schools commonly raise about generative AI in education. That includes teacher judgement, privacy, bias, over-reliance, student voice, writing time and transparency. Referencing external standards is not a claim of endorsement by the organisations named.
This page is informed by public policy, technical documentation and Edsthetic's own product safeguards. These links are provided for transparency; they do not imply endorsement of Writeiq by the organisations named.
This page is a statement by Edsthetic, the company that makes Writeiq. It is written to be checked against the product and against our Security and Privacy, Privacy policy and Terms of service pages, not taken on trust. If something here does not match your experience of the product, please tell us.