How student work is sent to the AI
DeepMarking marks with Google Gemini. Your students' files are opened in your browser, and only the extracted text, readable images and a list of file names are sent on for marking.
The path a submission takes
- Your browser opens the files. ZIPs are unpacked on your computer. Code and images are read there too.
- Text is extracted from documents. PDF and Word files are sent to a DeepMarking server that extracts the text and does not keep the file.
- The submission is queued. When you start marking, each student's text and images are held in private storage until that student is marked, then deleted. Anything left by a run that never finishes is deleted within a day or two.
- Gemini marks it. DeepMarking sends the marking request to Google Gemini and saves the marks it returns.
- The text is kept. The text the submission was marked from is stored with its marks, so you can re-mark it later.
What the marking request contains
- The spec text and your rubric's criteria.
- The submission's text and code, up to 220,000 characters.
- Up to 24 images.
- A list of the student's file names.
- The student's name and ID as DeepMarking detected them.
- Your strictness and assessment type settings.
- Lessons accepted from your own earlier reviews of the same assessment.
Which account is used
All AI calls use DeepMarking's own platform credentials. You cannot supply your own AI key.
What our Privacy Policy says
The Privacy Policy describes AI processing as sending the assessment text the model needs to return marks and feedback. It lists Google as a processor for Gemini. It makes no further statement about how Google uses that data. If your institution needs a specific answer on model training, email [support@deepmarking.com](mailto:support@deepmarking.com) before you upload.
Read the full Privacy Policy.
Your responsibilities
You are usually the controller of any student personal data you upload. Upload student work only when you have a lawful basis and your institution's permission. Do not upload content you are not permitted to process.
If your institution requires it, remove student names from documents before you upload. The name DeepMarking reads from file names or cover pages is still included in the marking request.
Related articles
- What data DeepMarking storesYour account details, your specs and rubrics, the text each submission was marked from, and the marks. Student original files stay on your computer.
- Supported file types and upload limitsWhich files the marker can read, which are listed by name only, and the upload limits: 2,000 files, 250 MB in total, and 20 MB per PDF or Word document.
- Who can see your projectsYour projects are private to your account. There is no sharing feature. DeepMarking administrators have access through admin tools. What that means for you.
Was this helpful? If not, .