Draft analysed
The AI reads the work against your rubric.
Upload the class, let it draft a rubric-aligned judgement for every student, then go through them one by one and decide. It does the reading. You do the marking.

ISMGenius by the numbers
Assignments Analysed
Students Supported
Academic Integrity Scans
Words of Feedback Generated
Every result is a suggestion, not a verdict.
It appears in your moderation workspace against each criterion, ready for your review. Nothing becomes a mark until you confirm it.
Our estimates are QCAA-aligned — not QCAA outcomes. We don't replace your expertise; we save you time.
Get a reliable first read on every draft in minutes, not hours, so you can give better feedback where it matters most.
The AI reads the work against your rubric.
Suggestions appear in your moderation workspace.
Confirm every mark and hand back with confidence.
ISMGenius gives you a consistent, QCAA-aligned first read so you can focus on what only you can do.
Create a class, then paste your roster in — one student per line, with an optional student ID. Add an assignment and upload its task sheet and ISMG. You do this once; every assessment for that class reuses it.
Upload up to 200 PDFs at once. Each file is matched to a student by the name or student ID in its filename. A full-name or ID match files itself; a single-token match is filed but flagged for you to confirm; anything ambiguous is held aside rather than guessed.
One click queues every submission that still needs marking. The run is durable — it finishes on the server whether or not your browser stays open, and picks up again if something fails partway.
Every result opens as a suggestion beside your judgement, criterion by criterion. Accept it, adjust it, or overrule it entirely. Nothing is a mark until you finalise it.
Print one page per student, or export the whole class as a gradebook CSV with your marks and comments. Students never need an account — you hand the feedback over yourself.
From whole-class marking through moderation to hand-back, every part of it works in one place.
Upload up to 200 PDFs in one go, matched to your roster by filename. Confident matches file themselves, uncertain ones are flagged, and anything ambiguous waits for you rather than being guessed.
Each criterion shows the AI's best-fit judgement next to yours. Fill in the suggestions and adjust, or mark from scratch. Nothing becomes a mark until you finalise it.
A cohort-level read on where the class landed: what most students did well, where the common shortfall was, and what is worth reteaching before the next task.
A class holds your roster and every assessment you set it. Paste a list to build it in one go, and use names or student numbers — whichever you would rather have in your gradebook.
Every student carries two tracks. Mark the draft, then the final against the same rubric, and see both on one line — including whether the work moved between them.
Optional AI-likelihood and source scanning, per submission or per class. Every score carries when it was run and how much of the document it covered.
One page per student. Print the class in a single pass or copy a single sheet into your LMS — the student needs no account.
Add a co-teacher who can mark alongside you, or a view-only colleague. Internal moderation is a multi-teacher process, so the toolkit treats it as one.
Student work is never used to train AI models and is never sold or shared. Deleting a student, assignment or class removes the files, the feedback and the scans with it.
Every one of these ships in the toolkit today.
Run it only when you want it, see exactly what it covered, and keep the conversation with the student on the record.
Detection runs only when you ask for it, on a submission or on a class. A teacher who would rather not use it never has to.
Long submissions exceed the detector's input limit, so a score can reflect part of a document. It always says which, and when it was run.
Each case holds what you checked, what the student said, and what happens next, so the discussion is what settles it.
No detector identifies AI writing correctly in every case. A percentage should never be the sole basis of a misconduct finding, and the toolkit is built to keep that limitation in front of you.
Apply with your school email. If your school's domain is recognised it is matched automatically; otherwise just tell us where you teach. There is no charge and no card.
Try it on a single draft first — no account needed. Then apply for the toolkit and bring the whole class.
How the criterion estimates are produced, and what to tell students who ask about AI use.