Objective sections, instantly
Multiple choice, fill in the blanks, true or false and match the following are scored the moment a paper is submitted, against the answer key set when the paper was drafted.
Instant AI Evaluation
Marked before they've left the room.
Objective sections are scored the moment a paper is submitted. Written answers are read against your rubric in seconds, point by point, with the reasoning shown beside every mark — and a teacher who can change any of it before a result is declared.
Objective
Written
Rubric-scored
Teacher override
Full audit trail
3.1 secondsAverage per script
Written answersNot just objectives
Reasoning shownOn every mark
TeacherHas the final say
What it marks
Six jobs, and a
Six jobs, and a
teacher over all of them.
Marking a tick-box is arithmetic. Everything below is the work of scoring an answer a person actually wrote, and being honest about how confident that score is.
Written answers, point by point
Short and long answers are read against the expected points, not matched to a phrase. A student who explains it in their own words gets the mark.
Your rubric, not ours
Attach the marking scheme you already use. Each expected point carries its own marks, and the evaluator reports against those rather than an internal scale.
Partial credit that makes sense
Three of five points earns three marks, and the script shows which two were missed. A half-right answer is never rounded to nothing to keep the arithmetic tidy.
Spelling and phrasing tolerance
A misspelled technical term still earns the point if the meaning is clear. You set how far that tolerance goes, per subject, and it applies consistently.
Re-marking on appeal
A student can query a score. The answer is re-evaluated, the teacher sees both the original and the new working side by side, and the outcome is logged either way.
Script → score
The working, shown
The working, shown
beside the answer.
One three-mark question, one student's answer, and every judgement the evaluator made in reaching a score. This panel is what a teacher sees — not a number on its own.
Marked in 3.1s
ANSWER SCRIPT
Half-Yearly Examination · 2026–27
Question as printed on the paper, followed by the candidate's response.
Question 17Short answer · 3
Q.
Explain why a gas fills its container completely while a liquid does
not, with reference to intermolecular forces.
Ch 5 · Moderate
3
Candidate's answeras submitted
A.
In gases the particles are very far apart so the force between them
is almost nothing. That is why they move freely in all directions
and spread out to fill whatever container they are put in. In
liquids the particals are closer and there is more force pulling
them together, so a liquid keeps its own volume and only takes the
shape of the bottom of the container.
Submitted 10:47 · autosaved 14 times during the exam
AI evaluation
High confidence
Expected points · from your rubric
Gas particles are far apart with negligible intermolecular force “very far apart so the force between them is almost nothing”
1 / 1Therefore a gas expands to occupy the whole container “spread out to fill whatever container they are put in”
1 / 1Liquid particles are held by stronger forces, so volume is fixed but shape is not Volume and shape both covered, but the force is described rather than named as intermolecular attraction
0.5 / 1Bonus: reference to kinetic energy overcoming attraction Not attempted — not required for full marks on this rubric
0 / 0
2.5
of 3
Spelling of “particals” ignored per subject setting. Teacher can change this score.
Accept
Override
The pipeline
Forty-two scripts,
Forty-two scripts,
one exam window.
Scripts are marked as they arrive rather than in a batch at the end, so the last student to submit isn't waiting on the first. The dashed stretch is the teacher review window — the only part that takes as long as a person wants it to.
123456
789101112
Roll 01submitted 10:41
Roll 07submitted 10:43
Roll 18submitted 10:45
Roll 24submitted 10:47
Roll 33submitted 10:52
Roll 42submitted 10:58
Objective sections scored
Written answers evaluated
Marked, held for the teacher to release
Illustrative timings, in seconds from submission. Nothing is published to a student on this track — scripts sit in the last state until a teacher releases the result, however long that takes.
Where it agrees, and where it doesn't
Ninety-four percent,
Ninety-four percent,
broken down honestly.
How often a teacher accepted the AI mark without changing it, by question type, across our own sampling. The aggregate is 94%. The two rows at the bottom are why teacher review still exists.
Question type
Marks a teacher accepted unchanged
Rate
Review
Multiple choice
100%
Not needed
True or false
100%
Not needed
Fill in the blanks
99%
Spot check
Match the following
98%
Spot check
Definitions
96%
Spot check
Short answer
93%
Review advised
Long answer
89%
Review advised
Measured on our own sample, not a customer's, and published because the shape matters more than the headline. Objective marking is arithmetic and behaves like it. Long-form marking is a judgement, and roughly one script in nine deserves a second look — which is the whole argument for the next two sections.
A score without a reason is just a number. Every mark traces back to a line in your rubric and a phrase in the candidate’s own answer — including the half marks.
The setting that decides the marks
How strict, and
How strict, and
who decides it.
One dial per subject, set by the department rather than by us. It moves the line between accepting a correct idea and requiring the correct terminology — and it is applied identically to every script in the division.
0
Agreement
Lenient accepts the idea
A student who explains osmosis correctly without ever using the word earns the mark. Right for lower standards and for the first assessment on a new chapter.
Strict requires the terminology
The expected term has to appear. Right for board preparation, where the real examiner will be looking for exactly that word and it does a student no favours to pretend otherwise.
The tick is our recommendation
Slightly toward strict, which is where teacher agreement peaked in our sampling. It is a default, not a lock — and moving it re-marks nothing that has already been released.
The gauge reads agreement at the current setting. Pushing all the way to either end costs you a few points of it, which is the honest trade rather than a hidden one.
Four modes, one setting
Not every paper
Not every paper
should be auto-marked.
Set per exam, not per platform. A weekly chapter test and a pre-board do not deserve the same amount of human attention, and pretending otherwise wastes a teacher's week.
Marked and released
Scores publish to students as soon as evaluation finishes. No teacher step at all, and the result is in their hands within seconds.
Reach for it whenIt's a practice set or a weekly chapter test and speed matters more than the last two percent.Sample, then release
The evaluator surfaces the scripts it was least confident about. A teacher reviews only those, then releases the division in one action.
Reach for it whenIt's a unit test with written sections and you want oversight without marking forty scripts.Every script seen
The AI mark becomes a first pass. A teacher opens each script with the working already laid out and confirms or changes it.
Reach for it whenIt's a term exam or a pre-board and the mark goes on a report card.Teacher marks it
Evaluation is switched off for the written sections entirely. Objectives are still scored automatically, because nobody wants to tick boxes by hand.
Reach for it whenThe question is genuinely open-ended, or the stakes make a human mark non-negotiable.
Questions
What teachers ask
What teachers ask
before they trust it.
The four that come up almost every time.
Against the expected points in your rubric, one at a time, rather than by comparing the answer to a model paragraph. Each point is judged present, partly present or absent, and the marks for that point are awarded accordingly — which is why a student who explains the idea in their own words still earns it. The working is attached to the script permanently, so a teacher, a parent or an examiner reviewing an appeal six months later sees exactly which points were credited and why.
A teacher overrides it, and that takes one click from the same screen where the working is shown. Roughly one long answer in nine gets changed, which is why long-form papers default to review rather than automatic release. Every override is recorded against the script and feeds our own sampling, so the honest answer is that it gets some wrong and the system is built around that rather than around a claim that it doesn't.
Yes. A student raises a query from their result, the answer is re-evaluated independently, and the teacher sees the original working and the new one side by side before deciding. The outcome is logged whichever way it goes, including when the original score stands. Schools tell us this is the part parents ask about first, and that being able to show the working is what ends the conversation.
No. Answer scripts are not used as training data, not shared with a third party, and not retained beyond the period your institute sets. Teacher overrides improve the rubric matching for your own institute only. Data residency and retention are agreed during onboarding and written into the configuration rather than only into a contract.
Ready when you are
Send us one answer script.
Book a free 30-minute demo. We'll mark it live, show the working, and you can tell us where we got it wrong.