September 29, 2026 · 7 min read
Checking an Assignment Before You Hand It In
The five-minute pass before you submit: file format, upload size limits, metadata, and reading your own draft critically — including how AI it sounds.
Most marks lost at the submission stage have nothing to do with the work. A file that wouldn't upload before the deadline, a document that opened blank on the marker's machine, a version that turned out to be the wrong draft. Here's a short pass to run before you press submit, in the order that saves you the most trouble.
1. Read the brief again, for the boring parts
- Required format. PDF and .docx are not interchangeable if the brief names one — some markers annotate in Word and some portals only accept PDF.
- Word count, and whether references and appendices count towards it.
- Filename convention. Many courses want a student number and the assignment code, and a marker with sixty files named essay-final.docx will not enjoy yours.
- Whether it must be anonymous. If so, your name should not be in the document, the header, or the file properties.
- Upload size limit. This is the one that ruins evenings, and it's usually written in small print on the submission page.
2. Get the file under the limit
Submission portals commonly cap uploads somewhere between 2 MB and 100 MB depending on the system and how it's configured, so check yours rather than guessing. If you're over, the cause is nearly always images — screenshots, diagrams, photographed lab notes or scanned pages — because text in a PDF weighs almost nothing.
- Export properly first: in Word or Google Docs, a straight Save as PDF is already small. Avoid "high quality print" settings.
- Resize screenshots and photos before placing them, rather than shrinking them in the document and leaving the full-size file embedded.
- Then compress the PDF if you're still over. Compressing the embedded images is what shrinks it; the text layer is not the problem.
- Do not export your essay as an image or a scan to make it smaller. Similarity checkers need real text, and an image-only PDF can come back as unreadable or unprocessable.
- Check the compressed version opens and that figures and tables are still legible before you consider it done.
3. Check what the file says about itself
Word and PDF files carry an author name, a title, a creation date and editing history. If you started from a friend's document, a template or a downloaded example, the author field may still name them — and on an anonymous submission that undoes the anonymity. Clear the author and title fields before submitting; on Windows it's right-click → Properties → Details → Remove Properties, and on a Mac you can check in the document's inspector.
4. Then actually reread it
- Answer the question that was asked. Reread the prompt and check each paragraph earns its place against it.
- Read one page aloud. You'll hear run-on sentences and missing words your eyes skip.
- Check every citation exists, matches the reference list, and that page numbers are real.
- Look at figures and tables: numbered, captioned, referred to in the text, legible at the size they print.
- Open the final file on a different device or in a different app. Fonts substitute, layouts shift, and equations break.
5. How does your own draft read?
If your course allowed you to use an AI tool for outlining, drafting or editing, and you've rewritten the result in your own words, it is reasonable to want to know how much of the original phrasing survived. Not because a score proves anything, but because heavily generic writing is weak writing: hedge-everything sentences, tidy three-item lists, the same connectives every paragraph, and confident claims with no source. Reading a draft with that lens improves it regardless of how it was written.
See how generic your phrasing reads before you hand it in.
Open the AI Text DetectorWhat a detector score is and is not
This part needs saying bluntly, because a lot of students have had a bad week over it. No detector — not this one, not GPTZero, not Turnitin's — can prove who or what wrote a text. They measure statistical patterns, and plenty of human writing matches those patterns: clear technical prose, formal academic register, and writing by people working in a second language are all flagged more often than they should be. A score is a prompt to look again at your own writing, not evidence about it.
- Treat a high score as "this reads generically", not "this looks like cheating".
- The fix is the same either way: add your own examples, your own data, your own argument, and specific detail only you would know.
- Don't run someone else's work through a detector and treat the number as a verdict. It isn't one.
- Never rewrite honest work purely to move a score. That's optimising for a bad measurement, and it usually makes the writing worse.
Follow your institution's rules on AI use, and if they're unclear, ask before you submit rather than after. Where a declaration is required, say what you used and how. That conversation is far easier before a mark exists.
6. Submit early enough to fix a problem
- Aim for a few hours before the deadline. Portals get slow and occasionally fall over when a whole cohort submits at once.
- Keep the confirmation email or receipt screen. It is your only proof.
- Download your own submission afterwards and open it. Uploads do occasionally truncate.
- Keep a dated copy of the exact file you submitted, and your working drafts. Version history is useful if anything is ever questioned.
What these tools actually measure, and where they get it wrong.
What AI detectors flag