HEC Plagiarism Policy: What Similarity Percentage Is Acceptable?

HEC's plagiarism policy applies two limits, and most Pakistani universities enforce both. Your overall similarity index should be 19 per cent or less, and no single source should contribute more than 5 per cent. Your own department usually sets a tighter figure than either. The number is a screening threshold, not a pass mark.

That last sentence is the one students skip, and skipping it is expensive. Every year people with a perfectly acceptable score still end up in front of a committee, because they read the front page of the report and stopped there.

The two limits

HEC's plagiarism policy is built around a similarity index, the proportion of your document that matches text already in the database. Two figures come out of it, and most Pakistani universities apply both. They come from HEC's own plagiarism policy, and the figures below are the ones in force as this is written in 2026 (check the current document if you are reading this later):

  • An overall similarity index of 19% or less. Above that, the document is normally referred for review rather than rejected outright.
  • No single source above 5%. This is the one students forget. A thesis can sit comfortably at 14% overall and still be referred, because eleven of those points came from one paper.

The HEC policy sets two limits, not one The HEC plagiarism policy applies two figures. The overall similarity index should be 19 per cent or less. No single source should contribute more than 5 per cent. The second limit is the one students forget: a thesis can sit at 14 per cent overall and still be referred for review, because 11 of those points came from one source. Many departments set a tighter figure than the national one. Two limits, not one 19% Overall similarity index, or less 5% Any one source, at most THE ONE STUDENTS FORGET A thesis at 14% overall can still be referred, if 11% of it came from one source. Your department may set less.

The second limit is the more revealing of the two. A document that borrows a little from forty sources looks like research. A document that borrows a lot from one looks like a rewrite of that one, and the policy is written to catch exactly that shape.

Your department's rule beats the national one

HEC sets the ceiling. Individual universities, and often individual departments, set a tighter one beneath it: 15% is common, and some supervisors will not sign off above 10%.

So before you spend a week rewriting to hit 19%, find out what your department actually requires. It is usually in the thesis handbook, and if it is not, your supervisor knows. Two minutes of asking beats a fortnight of guessing.

What the percentage actually measures

A similarity index is a text-matching score. It counts matched strings. It does not know whether a match is plagiarism.

This matters in both directions:

A high score is often innocent. Your references section matches every other references section that cites the same papers. Standard methodology phrasing, such as "a semi-structured interview protocol was administered to", matches thousands of theses because there are only so many ways to write it. Block quotes you have cited properly still count as matched text. A thesis with a long bibliography and a methods chapter can run past 20% without a single dishonest sentence in it.

A low score is not a clearance. Take someone else's argument, reorder the clauses, swap in synonyms, and the matcher may find nothing at all. The score will be beautiful. It is still plagiarism, and it is the kind a supervisor who knows the literature spots by reading, which is how most of these cases are actually caught.

The number is a place to start looking. It is not a verdict.

How to read your institution's report

This section is about the report your university runs, which is Turnitin in most Pakistani institutions. That is the one with a per-source breakdown and exclusion settings, and it is the one your result is judged on. If you want the longer version, what a Turnitin Similarity Report actually shows goes through every panel of it, and explains why two classes can score the same document differently.

Do not look at the front page. Open the matched sources and go down the list.

  1. Sort by percentage and look at the top three or four sources. This is where the single-source limit lives, and it is the fastest thing to check.
  2. Exclude the bibliography, then look again. Turnitin can do this; not every tool can. If a quarter of your score was the references list, you have a very different problem from the one the front page suggested.
  3. Read every match over about 2%. For each one, ask whether it is quoted and cited, standard phrasing, or something you need to rewrite.
  4. Look at the matches nobody flagged. The sources you leaned on hardest are the ones you are most likely to have paraphrased too closely, and they will not always be at the top of the list.

How to read a similarity report, in four steps Four steps for reading a similarity report. First, sort the sources by percentage and check the top three. Second, exclude the bibliography and look at the list again. Third, read every match over about 2 per cent. Fourth, check the sources you leaned on hardest. By the end you should be able to name every source in your top ten. Read the list, not the front page 1 2 3 4 Sort by percentage. Check the top three sources. Exclude the bibliography, then look again. Read every match over about 2 per cent. Check the sources you leaned on hardest. Name every source in your top ten.

By the end you should be able to name every source in the top ten and say in one sentence why each one is there. If you cannot, that is the work.

Two reports, read side by side

Neither of the reports below is a real student's. Both are built from the patterns above, and the pair shows why the front page misleads in both directions.

Take a thesis that comes back at 24 per cent. That number frightens people, and the source list usually explains it in about ten minutes. Suppose nine of those points are the reference list, which matches every other reference list citing the same papers. Four are the methods paragraph, phrased the only way that paragraph is ever phrased. Three are block quotes, set in quotation marks and cited. The largest single source sits at 3 per cent. Exclude the bibliography and the score reads 15 per cent, with nothing near the 5 per cent line. Nothing here needs rewriting. The 24 was a number nobody had read.

Now take one that comes back at 14 per cent. That clears the overall limit, which is the only limit most people remember, so the student stops looking. Suppose eleven of those fourteen points trace to one review article they leaned on through the whole literature chapter. Naming that article in the citation does not help as much as people expect, because what the matcher found was the wording: passages close enough to the original to need quotation marks they never carried. Eleven points from one source is the shape the 5 per cent limit exists to catch.

The two differ in the only way that matters. One carries a high number and a clean shape. The other carries a low number and the shape a reviewer is looking for. Read the shape, not the front page.

Bringing a score down honestly

Rewriting to beat a matcher is a bad use of a week and it produces worse writing. What genuinely lowers a score also improves the thesis:

  • Quote deliberately, and less. A long block quote is matched text and it is usually a sign the point has not been made in your own terms yet.
  • Paraphrase from your notes, not from the source. Read the paragraph, close it, and write what it said. Paraphrasing with the original on screen produces the near-copy the matcher is built to find.
  • Cite the borrowed idea, not just the borrowed sentence. This does not move the number much. It is the thing that decides how a committee reads the number.
  • Leave the references alone. They are supposed to match. Exclude them from the score if your tool allows it and stop worrying about them.

Check before you submit, not after

The single most useful thing here is timing. A similarity report is useful when there is still time to act on it: a week out, when a 24% can become a 12% with two afternoons of honest revision. The same report the night before submission tells you something you can no longer do anything about.

Run a check early enough that the answer is still actionable. Run it again after you revise, because a rewrite can introduce new matches as easily as it removes old ones. If you have not used an independent checker before, how it works is the four-step version: upload, pay, wait for the report, download it.

One honest caveat

If your university submits to Turnitin, that is the report that counts, and only your institution can produce it. Anyone selling you "a Turnitin report" is selling you something they cannot deliver.

What an independent check gives you is an early read: an overall similarity score, plus the matched passages and the source each one came from. It does not break the score down per source, and it has no bibliography-exclusion setting, so the two limits above are your institution's report to check, not this one. The database is different too, so the figure will not match theirs.

What carries across is the part that matters: which paragraphs are matching and why. Fix those and the institutional score follows.

That is what our plagiarism check is for: a look at your document while there is still time to change it. Pay per document, get the report, and go and fix the paragraphs it points at. Pricing is per document, with no account and no subscription, and each check is its own purchase, so budget for two if you plan to re-check after revising. If something about your file or your order is unclear, contact us before you pay.

Frequently asked questions

What similarity percentage is acceptable in Pakistan?

Most Pakistani universities apply HEC's two figures: an overall similarity index of 19 per cent or less, and no single source above 5 per cent. Many departments set something tighter, and 15 per cent is common. Work to your department's figure rather than the national one, because your department's is the number your supervisor applies.

Is 20% similarity too high for a thesis?

It depends entirely on where the 20 per cent came from. Twenty per cent spread across forty sources, with the bibliography and the quotations counted in, is an ordinary literature review. Twenty per cent where twelve of the points come from one paper is a referral, whatever the front page says. Open the source list before you decide.

Does the reference list count towards the similarity score?

Yes, unless somebody excludes it. A reference list matches every other reference list citing the same papers, so it can carry a large share of your score on its own. Turnitin has a setting for this, and the report itself explains who controls it. Look at your score with the bibliography both in and out.

What happens if my similarity score is too high?

Above the threshold a document is normally referred for review rather than rejected. A person reads the report and decides whether the matches are quotation, standard phrasing or copying. That is why the shape of a score matters more than its size. A referral is a conversation you can prepare for, so long as you have read your own report first.

Which plagiarism checker should I use?

That depends on how often you need one. A subscription is cheaper per check if you check every week, and a waste if you check a thesis once. Best plagiarism checkers compared sets out what each one costs and which model fits which reader.

Can I check my similarity score before I submit?

Your institution's report is the one you are judged on, and only your institution can run it. Ask whether your assignment allows resubmission, or whether Draft Coach is switched on for you. An independent check gives you an earlier read on which paragraphs match, and our FAQ covers how a check is run. It will not reproduce your university's figure.

Is this about AI detection as well?

No. A similarity score and an AI writing score are two different measurements, produced by two different systems, and a document can score low on one and high on the other. If your worry is AI rather than similarity, how to check if an assignment is AI-generated covers that question instead.