Grade the submissions
Closed questions grade themselves. On written answers, the AI suggests and you decide.
Evalmee grades the four closed question types automatically: true/false, single choice, MCQ and matching. The six open types stay graded by a human. On written answers, the Evalmee AI assistant suggests points and feedback, which you approve answer by answer before they count towards the grade.
What the Evalmee AI assistant does
The AI is never triggered on its own. You lock the grading rubric of the question, you click Generate feedback, you read it over, then you approve. Without approval, nothing enters the grade.
- Manual trigger: one button, on the question you are grading
- Suggestions, not grades: suggested points and written feedback
- Explicit approval: grading is recorded only by your click
- Anonymized data: no personal data is sent to the AI provider
Grading question by question, not submission by submission
The Grading page, By question tab, shows every participant's answer to the same question on a single page. You apply the same criterion forty times in a row instead of rebuilding it for each submission.
- Every answer to a question on one page: the criterion stays stable
- Less drift between the first and the last submission: that is the very point of this mode
- Grading submission by submission: also available, from the Grades tab
Splitting the grading between several graders
Once your colleagues are invited to the exam, the split is made by question: you tick one or more questions, you click the grader's name, and they inherit those questions.
- Assignment by question: tick, then click a grader
- Three sharing roles: Full control, Editor, Read-only
- Roles propagated: a role given on a folder applies to its subfolders and its exams
How does AI-assisted grading work?
You define the criteria and the scoring before the exam. After the exam, you lock the rubric of the question, the AI generates suggested points and feedback, you read them over and you click Approve. The grade only moves at that last step.
The order matters: the rubricRubricA grid crossing assessment criteria with performance levels described in full sentences, from the weakest to the strongest. The marker places the paper in a cell rather than awarding a figure: the mark becomes the consequence of a description shared in advance. is locked before the AI steps in, so the scoringMarking schemeThe rule that converts a candidate's answers into a mark. It sets the number of points per question, any penalties, and the way partial credit is awarded. Published before the exam, it constitutes the examiner's commitment as to what will be rewarded. is settled before any submission has been seen and applies to every answer.
The assistant is called Evalmee AI. It relies on OpenAI, processing takes place in the European Union, and the data sent is anonymized. Client personal data is never shared with an AI provider, nor used to train models.
| Step | Who acts | What happens |
|---|---|---|
| 1. Assessment criteria | You | You enter the criteria of the exam, one by one |
| 2. Rubric by question | You | You attach criteria to the question, write the grading instructions and set the points of each criterion |
| 3. Locking | You | The rubric is frozen for every answer to the question |
| 4. Generation | Evalmee AI | Suggested points and feedback, on explicit request |
| 5. Review | You | You keep, adjust or replace the suggestions |
| 6. Approval | You | The answer is graded, with the points assigned manually or by the AI |
What does the AI do, and what does it not do?
The AI suggests points and writes feedback on written answers. It does not grade closed questions, defines no criteria, validates no submission and does not detect plagiarism. Each of these four limits matches a design choice, not a temporary gap.
A product page that promises fully automatic grading describes an exam few institutions would agree to defend before a board. So here is the exact line. The comparison of online exam platforms puts the same question to the six other French vendors, with the public page each answer was read on.
| Task | Who does it |
|---|---|
| Grading a closed question: true/false, single choice, MCQ, matching | The platform, deterministically. No AI is involved |
| Defining the criteria and the points of the rubric | The grader, before the exam |
| Suggesting points on a written answer | Evalmee AI, on request |
| Writing feedback on a written answer | Evalmee AI, on request, or the grader |
| Grading a spreadsheet, a diagram, code, an oral answer or an uploaded file | The grader |
| Validating the grading of an answer | The grader, by clicking Approve |
| Detecting plagiarism or AI-generated text | Turnitin, a third-party service, with the client's licence |
| Proctoring the exam, detecting copy and paste and the second screen | The proctoring measures, without AI |
Evalmee supports anonymous grading of submissionsAnonymous markingThe removal of the candidate's identity at marking time, the paper carrying nothing but a number. The practice targets biases linked to name, gender, presumed origin or a student's reputation, whose effect on the mark awarded is documented by research.. Two further fairness levers come on top: question-by-question grading, which reduces the severity gap between the first and the last submission, and locking the rubric, which prevents changing the scoring mid-grading.
How does the scoring scheme guarantee fairness between submissions?
A locked grading rubric can no longer be changed without consequence. If you unlock it while answers have already been graded, that grading is reset: the product refuses to let one exam be graded under two scoring schemes.
The same principle applies to the paper: once the exam is open, the questions seen by participants can no longer be edited, so that everyone sat the same paper.
On choice questions, the scoring is explicit and configurable. The discordance mode, used in PACES, PASS and LAS, grades both the choices ticked and the choices left blank.
To see what a penalty does to a paper before settling on one, the test grade calculator applies the same arithmetic question by question, and prints the grading chart that follows from it.
The scoring settings of multiple-choice questions
- Reward: all or nothing, or proportional to the number of correct choices
- Penalty: none, cumulative or absolute
- Calculation base: standard, balanced, moderate or manual
- Floor: none, zero, or a value you set
- Discordance scoringBarème par discordance (discordance scoring)A way of marking multiple-choice questions, common in the French PASS, LAS and former PACES health tracks, where the mark depends on the number of discordances with the expected answer: an item ticked wrongly, or a correct item left blank. Zero discordances scores full marks, and each further discordance drops the mark one tier. Each faculty sets the tiers itself.: the grade depends on the number of gaps between the participant's answer and the expected answer
Can a scoring mistake be fixed after the exam?
Yes. From the Paper tab, the Edit button lets you revisit the correct choices, the weightings, the grading comment and the tags, then run Recalculate grades. Every grade in the exam is recalculated.
In practice: a question whose right answer had been ticked wrongly is fixed in two minutes, without running the exam again and without reworking submissions one by one. A question to discard comes down to a weighting reset to zero, followed by the same recalculation. The scoring of choice questions is changed the same way from the settings, even after participants have sat the exam.
One case remains: a grade that has to be reported on a scale other than the exam's point total. The score converter does the arithmetic and prints the full conversion chart, point by point.
Frequently asked questions
You define the criteria and the points before the exam, then you lock the rubric of the question. The Evalmee AI assistant then generates, on request, suggested points and feedback. You read them over and click Approve.
No. The four closed question types (true/false, single choice, MCQ, matching) are graded automatically, without AI. On written answers, the AI produces suggestions that the grader reads over then approves. It defines no criteria, validates no submission and does not detect plagiarism.
Grading assistance relies on OpenAI, with anonymized data and processing in the European Union. Client personal data is never shared with an AI provider, nor used to train models.
Yes, and it is the recommended mode. The Grading page, By question tab, shows every participant's answer to the same question. The aim is to reduce the severity gap between the first and the last submission graded.
Through Turnitin, which can be enabled in open-book mode and compares answers with online sources, with the answers of other participants and with AI-generated content. The service runs on your institution's Turnitin licence and only processes answers longer than twenty words.
Used by
- Teaching teams
- Exam boards
- Freelance graders
- Training centres
Grade your next session with the assistant
Start for free, no credit card required.