What DeepMarking does

DeepMarking marks to the rubric you give it, not to a generic idea of a good essay. Each criterion gets a mark out of its own maximum and a comment that says why.

The criteria can come from wherever you keep them:

• a rubric spreadsheet (XLSX, XLSM, XLTX, CSV or TSV);
• a rubric table inside a PDF or Word document;
• criteria pasted in as text;
• the criteria table printed in the assessment specification.

Choose how hard the marker is: Strict (the default), Balanced, Lenient, or Flag & review, which gives a tentative mark to any criterion without clear evidence and flags it for you. On every setting, full marks need specific evidence from the student's work.

The project also records what kind of work is being marked: written, code, design, media, data, practical or a mix. The marker looks for the evidence that kind of work needs: it traces the logic of code without running it, and reads what each screen of a design actually shows. A late penalty per day can be set on the project, and days late entered for each student are deducted from the total.

Drop in the assessment specification, the rubric and the students' work. The specification can be a PDF, a Word document, a text or Markdown file, or pasted text.

Student work can be:

• PDF, Word (.docx), plain text and Markdown files;
• source code in common languages, including Python, Java, JavaScript, TypeScript, C, C++, C#, Kotlin, Swift, PHP, SQL, HTML and CSS;
• screenshots and other images;
• ZIP archives, including ZIPs inside ZIPs.

A bulk download from Moodle or Canvas can go in as one ZIP. DeepMarking reads the file names the LMS gives each file, so a student's report, code and screenshots become one submission for that student. Blank copies of the marking sheet that the LMS adds are left out.

ZIPs are unpacked in your browser. One upload can hold up to 2,000 files and 250 MB once unpacked, and each PDF or Word document can be up to 20 MB. Files of other kinds are still listed for the marker, so it knows the work exists.

AI marks are a draft and you stay the marker of record. For each student you see every criterion with its mark and comment. Confirm it, or change the mark or the comment. Totals update as you edit.

Criteria the marker was unsure about are flagged with a question and suggested answers. Pick one, or write a note in your own words and the AI turns it into a final mark and comment. The submission counts as marked once nothing is left flagged.

You can also rate the quality of the marking for each student. Your confirmations and edits guide later marking of the same assessment. Student ID numbers and labelled name lines are stripped from those examples.

Confirming a submission's criteria earns bonus credits, up to 3 per submission. If you mark that submission again, the confirmations and the bonus are cleared.

DeepMarking reads the specification to tell group work from individual work, for example "in teams of three" or "must be completed individually". You can change the setting on the project.

Often one team member hands in the team's work and lists everyone on it. The other members are linked to that submission and marked on the team's files. Group criteria are marked once per run and every member gets the same mark. Individual criteria, such as a reflection or a contribution statement, are marked for each student on that student's own work.

• One Excel results sheet for the whole cohort.
• A filled marking sheet for each student, as Excel or PDF.
• A ZIP of every student's marking sheet, named so your LMS can match each file to its student.
• A CSV with the columns you choose: marks, comments, penalties and totals.
• Your own LMS grade export (CSV or tab-delimited text), with the grades filled in. Only the grade cells change; the rest of the file comes back as it was.

Marks stay in your account, so you can come back and export again after further edits.

Marking uses credits. One credit marks one submission: a report, a code project, or an app with its screenshots. A very large submission uses two. A submission where no work was found uses none, and drafting a rubric uses none.

The Free plan gives you one credit a day, which resets at midnight in your timezone, and one active assessment with one submission, enough to try it on real work. Pro, Plus and Max add a monthly pool of credits and higher limits, billed weekly, monthly, quarterly or yearly through Stripe.

Compare plans and prices

• Marking runs on Google Gemini through DeepMarking's own account. You cannot and do not need to add your own API key.
• When a batch starts, each student's extracted work waits in private storage until it is marked, then it is deleted. The text a submission was marked from is kept with its marks so you can mark it again.
• Product analytics never include student names or submission text, and you can turn analytics off in your settings.
• You can delete single projects, your marking history, or your whole account. Deleting the account removes your profile, projects, submissions and marks.
• Payments go through Stripe. DeepMarking does not store card numbers.

The Privacy Policy has the full detail. Schools and faculties can find answers for IT and privacy teams on the Institutions page.