16 September 2026 · 7 min read
The last hour before a submission is the worst time to discover a problem and the most common time to discover one. Almost everything on this list takes minutes to verify and costs real marks when it is missed. Work through it in order — the checks that might require action come first, and the cosmetic ones come last.
1. Run the similarity check with enough runway to act on it. The entire value of a pre-submission report is the gap between the result and the deadline; run the night before and you have a diagnosis with no treatment. A few days is enough for most assignments, longer for a thesis.
2. Read the AI writing indicator as well, where your institution produces one. The two scores are independent and a comfortable similarity figure tells you nothing about the other. If you are only watching the percentage, you are watching half the report.
3. Sort the flagged matches before changing anything. Correctly quoted and cited material is supposed to match and should be left alone. Standard terminology, instrument names and statutory wording are usually defensible. Only uncited text that tracks a source closely needs rewriting, and that group is normally much smaller than the headline figure implies.
4. Re-check after rewriting. A second report is the only way to know the rewriting worked, and finding out afterwards is not an option that exists.
5. Every in-text citation appears in the reference list, and every reference list entry is cited in the text. Mismatches in both directions are the most common referencing deduction in Nepali marking, and they are trivially checkable with a search.
6. Every direct quotation is inside quotation marks with a page number where the style requires one. Quoted material that is cited but not marked as a quotation is the most frequent unintentional plagiarism there is.
7. Paraphrased material is cited too. Putting something in your own words removes the quotation marks, not the citation — this is the point students most often get wrong in good faith.
8. Every reference is real and you have actually seen it. If a source arrived through a citation in another paper and you never opened it, either open it or cite it as secondary. Invented and unverified references are found more easily than students expect.
9. One referencing style, applied consistently, and the one your department named. APA and Vancouver and IEEE are not interchangeable, and mixing two inside one document is immediately visible in the reference list.
10. Re-read the actual question with the finished document open beside it. A well-written answer to a slightly different question is the highest-scoring way to fail an assignment, and it is invisible to you by the time you have been working on it for a fortnight.
11. Check the marking rubric line by line if you were given one. Every criterion should map to something a marker can find in your document; if a criterion has nothing behind it, you have located free marks with time still on the clock.
12. Confirm the word count against what was asked, and check whether your institution counts the reference list, appendices and tables. The convention varies, and a penalty for exceeding a limit you were inside on a different counting rule is a bad way to lose marks.
13. Front matter in your department's order — cover page, declaration, certificate pages, acknowledgement, abstract, contents, list of tables and figures. The order differs between institutions and it is checked more often than the writing is.
14. Then the file mechanics, which are where genuinely avoidable disasters happen: submit the correct final version, in the required format, named as your department specified, with tables and figures numbered and captioned, page numbers running correctly, and the document opening properly on a machine that is not yours. Export to PDF and read the PDF, not the source file — layout shifts between the two more often than people expect.
And submit early enough that a portal problem is survivable. Upload failures cluster in the final hour for obvious reasons, and a timestamp is not something you can negotiate with afterwards.
Related guides:
If you want the first four items done properly, send the draft with your deadline. You get both reports, the flagged passages explained, and an honest answer on whether anything actually needs doing.