You send your thesis to the library for the plagiarism check. The report comes back at 24%. Your submission date is three weeks away. It feels like a disaster, but it usually is not. In most cases a high similarity score comes from a few predictable places: the literature review, standard method descriptions, definitions, and text that should have been excluded from the check in the first place.
This guide shows you exactly how to bring your thesis below 10% without shortcuts that can land you in bigger trouble. The 10% figure comes from India’s UGC rules, so the examples lean that way, but the same method works whether your university uses Turnitin, iThenticate, DrillBit or another checker.
| Quick answer: To reduce thesis plagiarism to under 10%, follow these six moves: Run your draft through the same checker your university uses, chapter by chapter.Apply the exclusions your rules allow: bibliography, properly quoted text, small matches and standard equations.Rewrite the highest-matching sources first, which are usually in the literature review.Paraphrase with the read, close, write, compare method instead of swapping words.Quote and cite only the lines that must stay word for word.Recheck, aim for a buffer of 6 to 8%, and save the final report. Avoid synonym spinners, character tricks and paid “plagiarism removal” services. They create new problems instead of fixing the real one. |
Why 10%? The UGC rule in plain words
The 10% number is not random. The UGC (Promotion of Academic Integrity and Prevention of Plagiarism in Higher Educational Institutions) Regulations, 2018 split similarity into four levels. Anything up to 10% is treated as minor similarity with no penalty. Above that, the consequences get serious quickly.
| Level | Similarity | What happens to a thesis or dissertation | What happens to a faculty publication |
| Level 0 | Up to 10% | Minor similarity, no penalty | No penalty |
| Level 1 | Above 10% to 40% | Revised script within a set period, not more than 6 months | Manuscript withdrawn |
| Level 2 | Above 40% to 60% | Barred from submitting a revised script for 1 year | Withdrawal, loss of one annual increment, no new research students for 2 years |
| Level 3 | Above 60% | Registration for the programme cancelled | Withdrawal, loss of two increments, no new research students for 3 years |
Figure 1. The four UGC similarity levels for theses and dissertations.
Two things are worth knowing before you start. First, your university can set a stricter limit than UGC, so read your own plagiarism policy. Second, the regulation says the core of your work (abstract, hypothesis, observations, results, conclusions and recommendations) should be based on original ideas and should not have any similarities at all. So 10% is the ceiling for the whole document, not a budget you can spend on your results chapter.
Similarity score is not the same as plagiarism
This is the single most useful idea in this guide. Turnitin itself says its percentage is a similarity score, not a plagiarism score, and a high number can come from properly cited quotes, reference lists and common phrases. The tool only finds matching text. A person decides whether that match is a problem.
That changes how you should read your report. A thesis at 14% made up of 60 tiny matches of half a percent each is a very different case from a thesis at 9% where one source contributes 6%. The second one is the bigger risk, even though it is under the limit. Your job is to fix the meaningful matches, not to chase every highlighted word.
It also works the other way. A very low score does not prove a thesis is original. Checkers often cannot read text inside images or diagrams, and translated material may not show up at all. So the goal is honest, well-cited writing that happens to score low, not a low score by any means.
What you are allowed to exclude from the check
Clause 7 of the UGC regulations lists what similarity checks should leave out. Many students lose 4 to 8 percentage points simply because these items were never excluded when the report was generated.
| Item | Excluded under UGC? | What to do |
| References, bibliography, table of contents, preface, acknowledgements | Yes | Remove these pages from the file you upload, or ask the library to switch on the bibliography filter. |
| Quoted work reproduced with permission or attribution | Yes | Put it in quotation marks or a block quote with a citation, so the “exclude quotes” filter can catch it. |
| Generic terms, laws, standard symbols and standard equations | Yes | Cite the source and request exclusion for these specific matches. |
| Common knowledge or coincidental terms up to 14 consecutive words | Yes | Ask the operator to set the small-match filter in line with your university’s rule. |
| Your own published papers that form part of the thesis | Allowed by many universities, for example University of Delhi’s guidelines | Cite your paper, list it in the thesis and ask your supervisor to recommend exclusion in writing. |
| Abstract, hypothesis, results, conclusions, recommendations | No | These must be your own words. Rewrite any match here first. |
Before you start: what you need
| You need | Why it matters |
| Your final draft as an editable Word file | You will be rewriting sections, and PDFs slow that down. |
| Your university’s plagiarism policy | It tells you the limit, which tool is used and who can approve exclusions. |
| Access to the same checker the library uses | Different tools search different databases, so a free online checker’s score will not match the official one. |
| A reference manager such as Zotero or Mendeley | It stops citation gaps, which turn harmless matches into real plagiarism. |
| The PDFs of your most-used sources | You need to compare your paraphrase against the original. |
| At least 2 to 3 weeks before the deadline | Rewriting a literature review well takes time. Rushed rewrites tend to be patchwork. |
Step-by-step: how to reduce thesis plagiarism to under 10%
Step 1: Run a baseline check on the right tool
Start by finding out what your library actually uses. In India, the Ministry of Education’s ShodhShuddhi programme gives universities free access to plagiarism software, and since October 2023 INFLIBNET has provided DrillBit under it. Many universities also use Turnitin or iThenticate. Whatever the tool, check with that one, because each compares your text against a different set of sources.
Three practical tips for this first run:
- Check chapter by chapter. A single 20% figure tells you nothing. Chapter scores show you where to spend your time.
- Do not save practice drafts to the repository. If your draft is stored in the tool’s student repository, the final version can match against it later. The University of Delhi library advises setting Turnitin to “No repository” before checking a paper.
- Write down the baseline number for each chapter so you can see progress.
Step 2: Read the report by source, not by total
Open the source list and sort by percentage. In most reports, a small group of sources accounts for most of the score. Fix those first. Use this table to decide what each type of match needs:
| What you see in the report | What it usually means | What to do |
| One source at 3% or more | A copied or closely paraphrased block | Rewrite the whole passage from scratch (Step 5). |
| Dozens of matches under 1% each | Common academic phrases | Usually leave them. A small-match filter will remove most. |
| Matches inside your reference list | Bibliography was not excluded | Apply the bibliography exclusion (Step 3). |
| Match to your own published paper or earlier draft | Self-match | Cite it and request exclusion (Step 8). |
| Highlighted text that is already in quotation marks | Quote filter is off | Ask for the quote exclusion to be applied. |
| Standard definitions, equations, scale items | Generic or standard material | Cite, shorten, or move to an appendix (Step 7). |
Step 3: Apply the exclusions your rules allow
Most checkers have filters for bibliography, quoted text and small matches. In Turnitin’s newer report, the small-match filter defaults to 8 words, and quoted text inside quotation marks or block quotes can be filtered out. Many Indian universities set the small-match length to match the 14-word UGC rule for coincidental terms.
One catch: if you are a student using Turnitin, any exclusions you make yourself are temporary and do not change the official score your instructor sees. Official exclusions usually come from the library or your supervisor. Many universities, including the University of Delhi, let the supervisor recommend exclusions in a set format. Ask early.
Only exclude what the rules allow. Excluding a real source just to lower the number is the fastest way to fail a review panel.
Step 4: Rewrite the literature review first
The literature review is where you summarise other people’s work, so it is almost always the highest-matching chapter. The biggest fix is structural, not word by word: stop summarising author by author and start organising by theme.
| Author-by-author (high match risk) | Theme-based synthesis (low match risk) |
| Sharma (2019) found that remote employees reported higher productivity. Lee (2020) found that remote employees reported higher productivity but lower team cohesion. Patel (2021) found that productivity gains faded after six months. | Early studies agreed that working from home lifted self-reported productivity (Sharma, 2019; Lee, 2020), but the gains were not permanent. Lee (2020) noticed a cost to team cohesion, and Patel (2021) found the productivity boost disappeared within six months. Together these suggest the benefit is short-lived unless teams invest in staying connected. |
Notice what changed. The second version compares sources, groups them and adds your own conclusion. That conclusion is original thinking, and original thinking does not match anything in a database.
Step 5: Paraphrase properly with the read, close, write, compare method
Most high scores come from “patchwriting”, which means keeping the original sentence structure and swapping a few words. Checkers catch this easily. Use this five-step method instead:
- Read the passage until you can explain it to a friend.
- Close the source. Hide the tab or turn the page over.
- Write the idea in your own words, in your own sentence order, and connect it to your study.
- Compare with the original. If any run of words looks the same, rewrite that part again.
- Cite the source, even though the words are now yours. The idea still belongs to the author.
| Version | Text | Result |
| Sample source sentence | Small and medium enterprises face significant barriers to digital adoption, including limited capital, a shortage of skilled staff and concerns about data security. | Original |
| Weak paraphrase (patchwriting) | Small and medium businesses face major barriers to digital adoption, such as limited money, a lack of skilled employees and worries about data security. | Still matches; structure copied |
| Strong paraphrase | For smaller firms, going digital is held back less by interest than by resources: they often cannot fund it, cannot find people to run it, and worry about keeping customer data safe (Author, Year). | Low match, cited, adds a point |
Step 6: Quote only what has to stay word for word
Some text should not be paraphrased: a legal provision, a definition you are going to critique, a questionnaire item, or a line where the exact wording is your evidence. Quote these, but do it properly. Use quotation marks or a block quote plus a citation. The quote filter only recognises text that is actually inside quotation marks or formatted as a block quote, as Turnitin’s filter guide explains.
Keep direct quotes rare. A thesis built from long quotes may pass a filtered check but still read as thin to an examiner.
Step 7: Handle methods, definitions and standard text
Methods chapters repeat standard wording, because many studies use the same tests and procedures. You cannot invent a new name for a t-test, but you can still lower the match:
- Describe what you did, not the textbook procedure. Use past tense and your own details: sample size, location, dates, software version, any changes you made.
- Cite the standard protocol instead of copying its steps. One line such as “Data were cleaned following the procedure in Author (Year)” can replace a full copied paragraph.
- Move instrument items to a table or appendix and check with your library whether appendices are included in the scan.
- Leave standard equations alone. UGC excludes standard symbols and equations, so cite them and request the exclusion.
Step 8: Deal with self-plagiarism and your own published papers
If you have published parts of your thesis as journal papers, the checker will find them. That is normal. The fix is disclosure, not rewriting your own results into worse prose. Cite each paper where you use it, list your publications in the thesis, and ask your supervisor to request exclusion. Some guidelines, such as the University of Delhi’s, explicitly exclude published papers where the researcher is an author. Check your own rules before assuming the same applies.
Step 9: Close every citation gap
A missing citation does not change your percentage, but it changes what a match means. A matched sentence with a citation is borrowing. The same sentence without a citation is plagiarism. Go through each highlighted passage and confirm it has a reference. Do the same for figures, tables and data you adapted, since checkers often cannot read text inside images. A reference manager makes this far easier than fixing citations by hand at the end.
Step 10: Recheck, keep a buffer and save your records
Run the full thesis again with all approved exclusions applied. Aim for 6 to 8% rather than 9.9%, because a re-scan later, a database update or a different checker can move the number up a little.
Then keep your paperwork in order. The UGC regulations require you to sign an undertaking that the work is your own and checked through an approved tool, and your supervisor signs a certificate too. Under ShodhShuddhi, a one-page plagiarism report or certificate is uploaded with Shodhganga theses. Save the final report as a PDF, along with a short log of what you changed and which exclusions were approved.
Worked example: from 23% to 8%
Here is how the steps above typically add up. The numbers below are an illustrative example to show the order of work, not real data from a specific thesis.
Figure 2. Illustrative example. Your own numbers will differ.
| Stage | What was done | Change | Running score |
| Baseline | Full thesis uploaded with references, acknowledgements and quotes included | 23% | |
| Exclusions | Bibliography, quotes and small-match filter applied as allowed | -6 points | 17% |
| Literature review | Top 8 sources rewritten as themed synthesis | -5 points | 12% |
| Methods and definitions | Standard procedures cited instead of copied, scale items moved to appendix | -2.5 points | 9.5% |
| Quote and cite | Two legal definitions put in quotation marks with citations | -1.5 points | 8% |
Chapter by chapter: where similarity hides
| Chapter | Typical risk | Main fix |
| Introduction | Background facts copied from reports and websites | Summarise in your own words and cite. Cut generic background. |
| Literature review | Highest risk. Author-by-author summaries | Themed synthesis with your own comparisons (Step 4). |
| Methodology | Standard procedures and definitions | Describe your actual process and cite protocols (Step 7). |
| Results | Low risk, but must be fully original | Write findings directly from your own data. |
| Discussion | Medium risk when comparing with other studies | Lead with your finding, then cite the comparison briefly. |
| Conclusion and recommendations | Should have no similarity | Write fresh. Do not reuse lines from the abstract word for word if your tool flags them. |
| Appendices | Questionnaires and scales | Confirm whether appendices are scanned. Cite every instrument. |
Tricks that backfire (and why you should skip them)
Search for this topic and you will find plenty of hacks. They might move the number, but they turn a fixable writing problem into a possible misconduct case.
| Trick | Why it fails |
| Synonym spinners and article rewriters | They produce awkward, sometimes wrong sentences that examiners notice. The idea is still borrowed, and the meaning can drift. |
| Swapping letters for look-alike characters, or hidden spaces | Many checkers and reviewers look for unusual characters. If found, it looks deliberate, which is worse than a high score. |
| Pasting text as images | The copying is still there. It just hides from the software, not from your examiner or the viva. |
| Translating into another language and back | Translated copying is still plagiarism, and the result often reads badly. |
| Paying a “plagiarism removal” service | You sign an undertaking that the thesis is your own work. Someone else rewriting it can breach that. |
| Running chapters through an AI paraphraser | It may lower similarity but raise AI-writing flags, and you lose control of your own argument. See the next section. |
What about AI writing detection?
Many universities now run an AI-writing check alongside the similarity check. Turnitin reported that in the first year of its AI detector, over 200 million papers were reviewed, with more than 22 million showing at least 20% AI writing and over 6 million showing at least 80%. So using an AI tool to rewrite your thesis can trade one red flag for another.
AI detectors are also far from perfect. A Stanford study of seven AI detectors found they were near-perfect on essays by native English speakers but wrongly labelled 61% of essays by non-native speakers as AI-generated. That matters for many Indian researchers writing in their second language.
Figure 3. How often AI detectors wrongly flagged human-written TOEFL essays.
The safest approach is simple: write and rewrite yourself, keep your drafts and version history as proof of how your work developed, and follow your university’s rules on declaring any AI assistance.
Why getting this right matters after you graduate
A thesis does not disappear after the viva. Under UGC rules, universities must send theses to INFLIBNET’s Shodhganga repository within a month of the degree being awarded, so your work becomes publicly searchable. The same regulation says that if plagiarism is proved after a degree is awarded, the degree can be put in abeyance.
Published research faces the same scrutiny. A study in PNAS of 2,047 retracted biomedical papers found that 9.8% were retracted for plagiarism and 14.2% for duplicate publication. Clean writing in your thesis makes it much easier to turn chapters into papers later.
Figure 4. Reasons for retraction among 2,047 PubMed-indexed papers.
A simple timeline to stay under 10%
| When | What to do |
| 8 weeks before submission | Run chapter-wise baseline checks. List the top sources for each chapter. |
| 6 weeks before | Rewrite the literature review as themed synthesis. |
| 4 weeks before | Fix methods, definitions and citation gaps. Request exclusions for your own papers. |
| 2 weeks before | Run the full thesis with approved exclusions. Target 6 to 8%. |
| 1 week before | Final check, download the report, get supervisor sign-off and the certificate. |
Final thoughts
Getting a thesis under 10% is mostly about working in the right order. Check with the official tool, apply the exclusions you are entitled to, rewrite the few sources driving most of the score, and cite everything you borrow.
Choosing a suitable research topic is also important from the beginning. If you’re still planning your postgraduate research, this guide on how to choose a research topic for your M.Tech project can help you get started. Follow these steps, and the similarity score usually takes care of itself, leaving you with a thesis you can defend confidently in your viva and potentially develop into published papers.
Frequently asked questions
What is the acceptable plagiarism percentage for a PhD thesis in India?
Under the UGC 2018 regulations, similarity up to 10% is Level 0, with no penalty. Above 10% the student is asked to revise. Some universities set a stricter limit, so always check your own policy.
Does the 10% limit include references and the bibliography?
No. UGC Clause 7 excludes references, bibliography, table of contents, preface and acknowledgements. The exclusion still has to be applied in the software, so confirm it is switched on in your report.
Why is my score different on Turnitin, DrillBit and free online checkers?
Each tool compares your text with a different database of web pages, journals and student papers. Only the tool your university uses counts, so treat free checkers as rough practice at best.
Is a 0% similarity score good?
Not always. Turnitin’s own student guide notes that zero similarity can mean you have not included source-based evidence, which most academic work needs. Some matching is normal.
Can I use a paraphrasing tool to reduce plagiarism?
It is risky. Automated rewriting often distorts meaning, can trigger AI-writing flags, and does not change the fact that the idea came from someone else. Rewrite manually and cite the source.
Will my own published papers count as plagiarism?
The checker will match them, but many universities allow them to be excluded if you are an author and you cite them. Get your supervisor’s written recommendation for the exclusion.