16 September 2026 · 7 min read
Most Nepali students check a thesis once, on the assembled document, days before submission. By then a flagged literature review means rewriting a chapter that three later chapters are built on, with no time to do it. Staging the checks by chapter costs very little and removes the single worst surprise in the process.
A thesis is not one document as far as a similarity report is concerned. Different chapters produce different matches for structurally different reasons, and knowing which is which stops you panicking about a number that is doing exactly what it should.
The literature review is where the real work is. Paraphrase that stays too close to its source is the most common genuine similarity problem in a Nepali thesis, and it is almost never deliberate — it happens when you write with the source open beside you. The fix is to read the passage, close the source, and write what it said from memory, then cite it. Text produced that way is genuinely yours; text produced by swapping synonyms is structurally identical to the original and now carries a second problem.
Your supervisor's response time is the variable that decides your submission date — a fortnight per round of comments and three rounds on a chapter means six weeks of calendar time per chapter, however fast you write. Checks should sit inside that rhythm rather than being bolted on at the end.
That last full-document check matters more than students assume. Material carried over from your own approved proposal into the final thesis will match the proposal, and self-matching is treated seriously by some departments even though it feels like it should not count. Check it rather than discovering the department's view of it at submission.
Run the AI writing indicator by chapter too, where your institution produces one. The chapter distribution is informative on its own: a flag concentrated in your methodology is telling you something about the genre, since that chapter is formulaic by design and formulaic is what the detector responds to. A flag spread evenly across chapters is a different signal and worth taking more seriously as a prompt to check your own evidence.
Do not rewrite to move the score. There is no editing operation that reliably lowers an AI figure on honest work, and paraphrasing tools raise it — their output is generative and carries its own signature. What answers an AI flag on a thesis is the drafting trail, and a thesis is the one document where that trail is easy to build because it accumulates over months anyway.
Keep chapter drafts as separate dated files, keep your supervisor's comments and your responses, keep the notes and the abandoned sections. Nobody writes a thesis in one pass, and the evidence of that is worth more than any percentage.
Less than most students expect, and far less than a rewritten literature review in the final fortnight. We check chapter by chapter as chapters settle, issue the full report each time with the flagged passages classified, and tell you plainly which matches are doing their job and which need attention. That includes work we had no part in writing, which is most of what we see on theses.
The output is not a percentage. It is the report, the classification, and an honest answer on what to do — which on a well-referenced chapter is frequently nothing at all.
Related:
Send the chapter you are least sure about with your deadline, and we will tell you where it actually stands.