Can Ed.ai’s AI make mistakes or generate wrong content (“hallucinate”)?

Yes. Like any AI, Ed.ai’s AI can make mistakes. That’s why you check and approve every result: nothing reaches your students without your approval.

Transcription (reading the handwritten paper):

• it relies on an OCR model trained to produce a faithful digital twin of the paper;

• a reading error is still possible, for example a digit or a word misread in hard-to-read handwriting. If you spot one, report it to us before you go back over your papers (click your initials at the bottom left of the screen, then “Report a bug”, or email contact@ed.ai): we fix the reading and run the grading again.

See also Que faire si les copies sont illisibles du fait de la graphie de l’élève ?

Initial grading: we keep improving our models to make them more reliable:

• our technical and learning teams work together toward grading that is accurate, consistent and aligned with what teachers expect;

• your role stays central: you check, adjust and put the AI’s suggestions in context.

To learn more: AI Transparency Statement