ISMGenius for teachers

Two ways to use ISMGenius

One draft at a time. Upload a student draft alongside the task ISMG and get criterion-by-criterion feedback in seconds. This works on any account and needs no setup, which makes it the quickest way to try the marking quality before committing to anything.

A whole class at a time. The Teacher Toolkit, included with the Educator plan, adds class rosters, batch upload, whole-class generation, a moderation workspace and printable hand-back sheets. The rest of this page documents the toolkit.

Getting access

The Educator plan is granted by application rather than purchase. Sign in, go to For Educators and apply with your school email. If your school's email domain is recognised it is verified automatically; otherwise you enter your school name and an administrator reviews it. There is no charge and no card.

Classes and rosters

A class holds a roster and every assessment you set that group. Create one, then build the roster by pasting a list — one student per line, with an optional student ID after a comma:

Jonathan Lee
Priya Nair, 12B-12
Sam Okafor, 12B-19

A student here is a label you control, not an account. Nothing checks that it is a real name, so if you would rather no student name entered the system, use student numbers instead and every feature still works. Students cannot sign in and never receive anything automatically.

Assignments and the rubric

Add an assignment to the class, then upload its task sheet and ISMG once. Every submission for that assignment is marked against the same rubric. You can add marking instructions telling the marker what to be strict on or what the task constraints are. Confirm the rubric outline before processing, so you know what the marker extracted from the ISMG before it marks thirty students against it.

Batch upload and filename matching

Upload up to 200 PDFs at once. Each file is matched to a student on your roster by what is in its filename:

• The student's ID, or their full name — filed automatically.
• A single name token (a shared first name, a common surname) — filed, but flagged for you to confirm.
• A filename that could name two students on the roster — held as unmatched rather than guessed.

Anything unmatched waits in the submissions list for you to assign by hand, and any submission can be reassigned at any time. This conservatism is deliberate: attaching a mark and written feedback to the wrong student is the failure the matcher exists to avoid.

Generating feedback for the class

Generate all queues every submission that still needs marking. The run is durable — it is queued and completed on the server, so it finishes whether or not your browser stays open, and it resumes if an individual submission fails. You can watch progress or close the tab and come back.

Whole-class generation covers the draft track, which is formative and returns next steps a student could act on. Final feedback is generated one student at a time and assesses the completed work without prescribing revisions. Every student row carries both tracks separately, so a draft mark is never mistaken for a final one.

Moderating — confirming the marks

This is the step that matters. Each result opens with the AI's best-fit judgement beside an empty field for yours, criterion by criterion. You can fill in the suggestions and adjust from there, or mark from scratch and ignore them. Nothing counts as a mark until you finalise it, and an unconfirmed mark shows in muted italics everywhere it appears so you can always tell which marks in a class are yours.

Each submission also carries two separate pieces of writing:

• A teacher note, private to you and your records.
• A student comment, drafted for the student and yours to edit before it goes anywhere.

If new AI feedback is generated after you opened the workspace, the moderation is marked stale rather than silently overwritten, so you review the new result before saving over your judgement.

Academic integrity

Integrity scanning is optional and never automatic. You can scan a single submission or a whole class for AI likelihood and source matching. Every score records when it ran and how much of the document it covered — the detector has an input limit, so a long research task may be only partly scanned, and the view says so.

Where a reading is worth pursuing, open a case against that student and record what you checked, what the student said, and what happens next. Treat scores as a prompt for a closer human read. No detector identifies AI writing correctly in every instance, and a percentage should never be the sole basis of a misconduct finding. See Academic Integrity for the student-facing explainer.

Handing feedback back

Students have no accounts and receive nothing automatically, so hand-back is something you do. Two routes:

• Feedback sheets — one page per student, print the class in a single pass, save as PDF, or copy a single sheet into your LMS.
• Gradebook CSV — one row per student, with your confirmed marks in their own columns and the AI suggestions in others, plus integrity readings for anything you scanned.

Sharing a class

Add a co-teacher, who can mark and moderate alongside you, or a view-only colleague who can see the class without changing it — useful for internal moderation or a head of department checking consistency. Only the class owner can rename it, delete it, or manage who else has access. Work stays attributed to the class, so it survives a collaborator later being removed.

Whole-class summary

Once a class has feedback, the summary reports how many drafts are in, the spread of bands per criterion, and a written cohort narrative built from de-identified feedback — student names are never sent to it, and it describes the class as a whole rather than singling anyone out.

Student data

• Student work is never used to train AI models, and is never sold or shared.
• Deleting a student, assignment or class also deletes the files, feedback, cached results and scans.
• There is no student-facing surface, so no student logins or contact details exist.
• Schools can request a data processing agreement and privacy impact assessment.

See Privacy for the full policy, or For Educators to get set up.

Accuracy and your judgement

ISMGenius produces QCAA-aligned estimates that support teacher assessment rather than replace it. They are not QCAA outcomes, the product says so wherever a mark appears, and you confirm every mark that counts.