Plagiarism and AI Rules for Indian PhD Scholars: What UGC, Your University and the Software Actually Say
Your thesis comes back from the check with two numbers on it: a similarity percentage and, increasingly, an AI-detection score. Neither number is a verdict. But they get treated like one, and most guides never explain who sets the rules behind each.
This guide separates three things that are constantly mixed up: what UGC’s documents say, what your own university requires, and what checking software merely indicates. It also shows how to read a similarity report line by line, and what to do about reusing your own published work.
Quick answer Three different things get mixed up. (1) UGC’s 2018 regulation sets similarity levels (up to 10%, 10–40%, 40–60%, above 60%) counted after standard exclusions, and concerns go through academic integrity panels. (2) Your university adds its own rules, which may include an AI declaration or AI limit. (3) Software scores, both similarity and AI-detection, are indicators, not findings. We found no UGC notification that defines AI-generated text as plagiarism or sets an AI percentage. The one UGC document we reviewed that mentions AI asks whether a journal has an AI policy. |
In this guide
▪ Similarity match vs confirmed plagiarism
▪ The four UGC similarity levels
▪ How to read a similarity report: a worked example
▪ Self-plagiarism for PhD scholars
▪ Who actually checks your thesis in India
▪ Where AI fits: three layers
▪ Using AI without putting your thesis at risk
▪ Pre-submission checklist and FAQ
Difference between Similarity match and confirmed plagiarism?
A similarity index is the share of your text that matches other sources in the software’s database. It flags overlap. It cannot tell you whether that overlap is a properly cited quotation, a reference list, a stock phrase, your own earlier paper, or copying without credit. Confirmed plagiarism is a finding, made by people who look at the matches in context.
| Similarity match | Confirmed plagiarism |
What it is | A software output: a percentage plus highlighted passages | A finding that words or ideas were presented as your own without acknowledgement |
Who produces it | The checking tool (for example DrillBit or Turnitin) | Your supervisor and, if a concern is raised, your institution’s academic integrity panel |
What it tells you | Where to look | Whether there is a problem, and how serious |
UGC’s 2018 regulation builds this distinction in. It excludes attributed quotations, references, generic terms and short common phrases from the count, and it routes concerns about a flagged report through a department-level panel and then an institution-level panel.
It works in the other direction too. The draft UGC PhD regulations we reviewed say an originality report is not sufficient proof that a thesis is free of plagiarism, and that a supervisor who suspects plagiarism can ask for an investigation. The same draft says only deliberate or grossly negligent infringement counts as research misconduct; not every breach of good practice does. So a high score is not a conviction, and a low score is not a clearance.
The four UGC similarity levels
For theses, the 2018 regulation quantifies plagiarism in four levels, based on the similarity index after exclusions:
Level 0 | Up to 10% | Minor similarity. No penalty. |
Level 1 | Above 10% to 40% | The scholar is asked to submit a revised script, typically within six months. |
Level 2 | Above 40% to 60% | The scholar is barred from submitting a revised script for one year. |
Level 3 | Above 60% | Registration for the programme is cancelled. |
What is excluded from the count: quoted work with proper attribution; references, bibliography, table of contents, preface and acknowledgements; generic terms, laws, standard symbols and equations; and common-knowledge or coincidental wording of up to 14 consecutive words. Universities can add their own rules on top of these national levels, so check yours.
How to read a similarity report: a worked example
The numbers below are invented for illustration. They show one thesis chapter with a raw similarity score of 22.0%.
What the software highlights | Share | What it really is | What to do |
Reference list entries | 5.1% | Excluded category | Nothing. Just confirm the report was run with references excluded. |
Quoted sentences with quotation marks and citations | 2.3% | Excluded if attributed | Check that every quote has marks and a citation. |
Standard method phrasing (“samples were centrifuged at…”) | 3.0% | Common phrasing | Usually excluded up to 14 consecutive words. Rewrite if it forms a longer block. |
Methods section reused from your own published paper | 6.4% | Self-reuse | Cite it, disclose it to your supervisor, and follow your university’s procedure. |
A paragraph closely following a review article, no citation | 5.2% | A genuine problem | Rewrite from your own understanding and cite the source. |
Raw score: 22.0%. After the three excluded categories are removed: 11.6%. Row four then decides the outcome. Where your university accepts a self-plagiarism exclusion, only 5.2% remains, which is Level 0. Where it does not, 11.6% sits in Level 1. Same report, different outcome, which is why you should learn your university’s procedure before you submit, not after.
A reading routine that works:
▪ Ask for the report with exclusions applied, not just the headline percentage.
▪ Sort every highlighted passage into one of the five buckets above.
▪ Fix genuine matches first: rewrite from understanding, then cite.
▪ Document each reuse of your own work, with your supervisor’s sign-off.
▪ Re-run the check on the same tool your university uses.
Self-plagiarism for PhD scholars: what needs citation, what needs disclosure
Three different things get blurred here. Citation credits the earlier work. Disclosure tells your supervisor, committee or journal editor that the overlap exists. Permission covers text or figures whose rights sit with a publisher. Reusing your own words is not automatically wrong. Doing it silently is the problem.
Common PhD situation | What to do | Watch out for |
A published paper from your PhD work becomes a thesis chapter | Cite the paper, say in the chapter that it is based on it, and follow your university’s procedure. Some universities use a supervisor-issued self-plagiarism exclusion certificate listing the published articles. | The procedure is set by your institution, not by one national rule. If the paper is co-authored, state your own contribution. |
Conference paper, then journal article, then thesis | Cite each earlier version, tell the journal editor about prior publication and keep the overlap limited. | Check the publisher’s terms for reusing text and figures. |
Figures or tables from your own paper | Mark them as reproduced from the paper and get permission where required. | Copyright often sits with the publisher, not you. |
Text from your MPhil or master’s dissertation | Ask your research cell before reusing it, and acknowledge the earlier work. | The draft UGC PhD regulations expect a certificate that the work has not been submitted for another degree, so earlier degree text needs particular care. |
The same background section across several papers | Rewrite it for each paper and cite your earlier work. | Journals run their own similarity checks. |
Who actually checks your thesis in India?
Most Indian institutions get plagiarism-detection software through INFLIBNET’s ShodhShuddhi programme. It provided Ouriginal (formerly Urkund) from 2019 until that contract ended on 30 September 2023, and has provided DrillBit-Extreme since 1 October 2023, to more than a thousand institutions. Scholars log in through their institution. Some universities also use Turnitin or other tools.
Tools use different databases and do not publish their full lists of indexed sources, so a number from one tool will not match another’s. A pre-check on a personal account or an older tool may differ from what your university reports. Ask your library or research cell which tool and which exclusion settings apply.
The draft UGC PhD regulations we reviewed also expect a thesis to go forward with an originality report, a scholar’s undertaking and a supervisor’s certificate of originality, and expect the final thesis to be deposited with INFLIBNET (Shodhganga), where it becomes openly accessible.
Where does AI fit? Three layers with different levels of certainty
Most confusion comes from treating these three layers as one. As of September 2026:
Layer | What it is | What we could verify | What to do |
1. UGC (national) | Regulations and public notices issued by UGC | The 2018 plagiarism regulation deals with text similarity and does not mention AI. The draft 2022 PhD regulations mention “other forms of academic dishonesty” but not AI. UGC’s February 2025 public notice lists an “AI generated content policy” as a suggestive criterion for choosing a journal. It is not a rule for theses. We found no UGC notification defining AI text as plagiarism or setting an AI percentage. Some websites claim otherwise without citing one. | Check UGC’s circulars page. Treat unsourced claims with caution. |
2. Your university | PhD ordinance, thesis format, declaration wording, any AI policy | Varies widely. Some institutions have published AI policies, for example NLU Delhi’s policy on AI-generated content in its journals and publications. | Ask your research cell for the AI clause and declaration wording in writing. |
3. Software indicators | Similarity and AI-detection scores from tools such as DrillBit or Turnitin | Indicators, not findings. Independent research shows detectors can wrongly flag human writing, especially by non-native English writers. | Treat a high score as a reason for human review. Keep dated drafts as evidence. |
Why AI scores need human review. A Stanford-led study published in Patterns tested seven AI detectors and found they misclassified more than half of essays by non-native English writers as AI-generated (a mean false-positive rate of 61.3%), while being near-perfect on essays by US students. Vendors dispute how far this generalises. Turnitin has said its false-positive rate is under 1%, while a small Washington Post test found a much higher rate. Many Indian scholars write in English as a second or third language, which is why an AI score alone should never settle a case.
None of this makes AI use harmless. Fabricated references and unverified claims are the real risks, and they are your responsibility.
Using AI without putting your thesis at risk
▪ Check your university’s AI clause and your target journal’s AI policy before you use any tool. UGC’s own journal-selection parameters list an AI policy as something to look for.
▪ Keep AI to language support (grammar, clarity) unless your university explicitly allows more. Anything that shapes your arguments, sources or data needs clear permission.
▪ Verify every reference yourself. AI tools can produce citations that do not exist.
▪ Do not upload unpublished data or manuscripts to public AI tools without checking their terms.
▪ Keep dated drafts and a short log of which tools you used and for what.
▪ Disclose in the format your university or journal asks for.
Pre-submission checklist
☐ Requested the report with exclusions applied, from the tool my university uses
☐ Sorted every highlighted passage: excluded, quoted, common phrase, self-reuse or genuine match
☐ Rewrote and cited every genuine uncited match
☐ Documented every reuse of my own published work and got my supervisor’s sign-off
☐ Checked quotation marks on every direct quote
☐ Confirmed my university’s AI clause and declaration wording in writing
☐ Verified that every reference exists and says what I claim
☐ Kept dated drafts in case a score is questioned
⬇ DOWNLOAD: PRE-SUBMISSION PLAGIARISM & AI CHECKLIST (PDF) The full checklist, with a similarity-report reading guide, a self-reuse log and a sample AI disclosure, as a free printable PDF. |
FAQ
Is a high similarity percentage the same as plagiarism?
No. It shows how much of your text matches other sources, including references, quotations and common phrases. Under UGC’s 2018 regulation, attributed quotations, references and short common phrases are excluded, and concerns go through institutional review, so a percentage alone does not establish plagiarism.
Does a low similarity score prove my thesis is original?
No. The draft UGC PhD regulations say an originality report is not sufficient proof that a thesis is free of plagiarism. Closely paraphrased ideas or uncited sources can still score low.
Is AI-generated text plagiarism under UGC rules?
We found no UGC notification that defines AI-generated text as plagiarism or sets an AI limit. The 2018 regulation covers text similarity. Your university’s ordinance or declaration may cover AI, and undisclosed AI use may conflict with the originality undertaking you sign, so check your institution’s clause.
Is there an official UGC percentage limit for AI-generated content?
We could not find one. UGC’s similarity levels (10%, 40%, 60%) apply to text similarity, not AI scores. Some universities set their own internal AI thresholds, so ask yours.
Can I reuse my published paper in my thesis?
Often yes, if you cite it, disclose it to your supervisor and follow your university’s procedure. Some universities use a supervisor-issued self-plagiarism exclusion certificate. Check the publisher’s terms before reusing text or figures.
Which plagiarism software do Indian universities use?
Many institutions use DrillBit-Extreme, provided through INFLIBNET’s ShodhShuddhi programme since October 2023; before that the programme provided Ouriginal. Some universities also use Turnitin. Ask your library which tool checks your thesis.
Are AI detectors reliable for non-native English writing?
Not reliably. A Stanford-led study found seven detectors wrongly labelled more than half of non-native English writers’ essays as AI-generated (a 61.3% mean false-positive rate). An AI score should prompt human review, not a conclusion.
Before you submit
Most problems are not deliberate. They are a rushed paraphrase, a reused paragraph nobody disclosed, or an AI-suggested reference that turned out not to exist. A second, trained pair of eyes before your official check can catch these while they are still easy to fix.
RE4U’s academic editing service reviews structure, paraphrasing and citation consistency alongside proofreading. We cannot tell you what your university’s software will report, and we do not promise any similarity score. Only your institution’s tool can do that.
Related reading: Understanding Plagiarism and AI in Research Writing · How to Spot a Predatory Journal Before You Submit
Sources and how current this is
-UGC (Promotion of Academic Integrity and Prevention of Plagiarism in Higher Educational Institutions) Regulations, 2018.
-UGC (Minimum Standards and Procedures for Award of Ph.D. Degree) Regulations, 2022, draft text reviewed (originality report, undertaking and certificate, ethics explanation, INFLIBNET deposit).
-UGC public notice F.No. 1-1/2018(CARE/JOURNAL)-Part file, 11 February 2025, and its suggestive parameters for peer-reviewed journals.
-INFLIBNET Centre, ShodhShuddhi programme brochure and portal.
-Liang et al., “GPT detectors are biased against non-native English writers,” Patterns, July 2023.
-National Law University Delhi, AI-Generated Content Policy.
Last verified: September 2026. Rules and policies on AI are changing quickly; confirm your own university’s current position before you submit. This article is educational and is not legal or institutional compliance advice.