16 September 2026 · 7 min read
It is the most frequently asked question about Turnitin in Nepal and the one with the least satisfying answer: there is no national similarity threshold. Nepal has no single figure that applies across universities, and any number presented to you as the national standard is somebody's departmental rule that travelled.
Thresholds in circulation across Nepali institutions vary widely. Some affiliated campuses work to around 20%, some postgraduate programmes are stricter, and a substantial number of supervisors set no figure at all and judge the report instead. All of those are legitimate positions, which is precisely why no single number can be quoted to you honestly.
The practical consequence is that the only threshold that matters to you is your department's, and the only reliable source for it is your department. Ask directly, and get the answer in writing if you can. A figure a senior passed down two batches ago is not policy, and departments do revise these rules — usually downward as the software becomes more familiar.
Be especially careful with numbers that cross institutions. A limit your cousin's college applies has no bearing on yours, and the confident specificity with which these figures circulate between students is not evidence of anything.
Before treating any figure as meaningful, find out what the report was configured to exclude. A report that counts the bibliography and the block quotations will read dramatically higher than one set to exclude them — on the identical submission. Two students can compare scores on the same file and get different numbers for no reason other than settings.
Reference lists are the biggest single distortion. A bibliography matches every other document citing the same sources, which is exactly what a correct bibliography should do. Many departments exclude it from the score; where yours does not, the report shows plainly that the match is bibliographic, and that is a defensible thing to point at.
Small quotation thresholds matter too. A report set to ignore matches under a certain word count reads lower than one counting every three-word run, and methodology sections in particular are full of standard phrasing that no discipline can avoid rewording.
Two submissions at 18% can be in completely different positions. One is spread thinly across dozens of correctly cited sources plus a reference list doing its job. The other is concentrated in three paragraphs tracking one uncited source. The first is fine and the second is a referral, and the percentage cannot tell them apart.
The inverse case is the dangerous one, and it catches students who trust the number. A 9% score sitting in one unbroken block of somebody else's argument is a far worse position than 22% made up of quoted material and a properly cited methodology. Low similarity is not the same as clear, which is the single most useful thing to understand about the metric.
So when you get a report back, sort the matches before you change a word. Correctly quoted and cited material — leave it; it is supposed to match. Standard terminology, instrument names and statutory wording your discipline cannot phrase otherwise — usually defensible and often excludable. Uncited text tracking a source closely — that is the group that needs work, and it is normally much smaller than the headline percentage suggests.
A properly referenced academic document will always match something: your own reference list, standard definitions, the names of instruments and Acts, and any material you have correctly quoted. A 0% score on a referenced document is closer to a warning sign than a success — it usually means quotation marks are missing where they should be, or the reference list is absent.
Treat anyone promising a guaranteed percentage with suspicion. The figure depends on report settings, on what else is in the corpus on the day, and on decisions your department makes. Nobody controls that output, so nobody can honestly guarantee it.
The target is a score your department accepts sitting behind a report that explains itself — not zero, and not a number you arrived at by deleting citations.
The AI writing indicator is reported separately from similarity and Nepali departments are increasingly reading both. A submission can show a comfortable similarity figure and still be referred on the AI score, so a student watching only the percentage is watching half the report.
The two also have opposite remedies. High similarity is fixed by citing, quoting and rewriting properly. A high AI flag is not fixed by rewriting at all — paraphrasing tools typically raise it, because their output carries its own signature. What answers an AI flag is a drafting trail, not an edit.
Related reading:
If you have a report you do not know how to read, send it with the document. We will tell you which matches matter, which are doing exactly what they should, and whether anything genuinely needs rewriting before you submit.