De-identifies before it scores
A name, year level and school together identify a student. Assessor strips that out before any work reaches the AI, and shows you exactly what it removed so you can attest that it did.
Assessor · AI marking for teachers
Mark a whole class set in one pass. Stay the marker.
Load a rubric, the class set and your roster. Assessor de-identifies the work, scores it against your rubric, says plainly what it can't judge, and hands back a live marksheet you review and own.
Live now. Load your first rubric and class set. No waitlist.
See it
The real thing, not a mock-up.
Rubric in, names stripped, class set marked. Every band traces back to the work, and you stay the one who signs it off.
Honest by design
Marking a teacher can defend, not a black box.
Declares what it can't mark
A criterion that needs evidence Assessor wasn't given (drafting history, an oral presentation, effort over time) comes back flagged as 'not assessable', with a reason. Never a fabricated score.
Every band cites the work
Each score points to a verbatim span from the submission. A band with no evidence is rejected before you ever see it. That's the guard against over-generous marking.
You are the final marker
Assessor produces a first pass. It never finalises a mark. You fill the flagged blanks, edit comments into your voice, override anything, and sign off.
The flow
From a folder of work to a marksheet, in five steps.
Upload
Sign in, upload your rubric, the whole class set of work, and your roster of names. Everything stays in the session only.
Approve the de-identification
Assessor detects every identifier: names seeded from your exact roster, plus school, year and any self-ID in the writing. You review and confirm what's removed.
Confirm the scope
It reads the rubric and proposes which criteria it can mark and which it can't from the evidence you gave it. You confirm.
Score against the rubric
Each de-identified piece is scored against your rubric with a band, an evidence span, and a positive, student-first comment.
Review and export
Names are reapplied locally, never sent to the model. You get a live Excel (real =SUM totals, flagged blanks, one row per student) to review and paste into Compass.
What you get
A live marksheet, not a static dump.
One row per student: separate name and topic columns you manage, per-criterion bands, a real =SUM total out of 25, a live percentage, and a four-sentence positive comment.
Criteria it couldn't mark are left blank and visibly flagged, so nobody copies a partial total by accident. Fill the blanks, and the total and percentage complete themselves, ready to paste straight into Compass.
FAQ
Assessor, answered
- Does student work go to the AI with names on it?
- No. Assessor de-identifies every submission before anything reaches the model, seeding the pass with your exact class roster so known names are caught deterministically. You approve what's removed before it proceeds.
- Does it give the final mark?
- No. Assessor produces a first pass and you are always the final marker. It fills a live marksheet you review, edit and sign off; it never finalises a mark on its own.
- What happens to the student work afterwards?
- Nothing is kept. Work, scores, comments and the re-identification map live in the session only and are gone when it ends. The one lasting artifact is the Excel you export.
- What if it can't judge a criterion?
- It says so. Criteria needing evidence it wasn't given (drafting, oral work, collaboration, effort over time) are returned as 'not assessable' with a reason and left for you, rather than scored as a silent zero.
- Does it detect AI-written work?
- It flags texture worth a look and suggests a verbal check with the student. It never emits a percentage-confidence 'AI detection' number.
Get your marking evenings back.
Assessor is live now. Load a rubric and your class set, and mark your first batch today. It runs on Year 7 English first, with more subjects on the way.