UGC Plagiarism Limits
The four similarity levels and the penalties attached to each, taken from the 2018 regulations themselves — including the exclusions most students never apply.
Why the Numbers Get Misquoted
Your supervisor quotes a percentage, a blog quotes a different one, and the consequences range from a resubmission to a cancelled registration. Almost every version online is a paraphrase of a paraphrase.
The Four Levels, From the Gazette
The regulations set four bands and say exactly what happens at each. They also exclude more from the count than most students realise, and they say nothing whatsoever about AI.
How it works
- Check which band your similarity score falls into
- Subtract what the regulations exclude — quotes, bibliography, standard terms
- Check whether your document is even covered; coursework is not
- Separate your similarity score from any AI score, which has no notified threshold
- Ask your IAIP for the figure in writing before acting on a number from a forum
The four levels and what each costs
Section 8 defines the bands; section 12.1 attaches the penalties for a thesis or dissertation. These are quoted from the regulation rather than summarised.
| Level | Similarity | Penalty for a thesis or dissertation |
|---|---|---|
| Level 0 | Up to 10% | Minor similarities, no penalty |
| Level 1 | Above 10% to 40% | Submit a revised script within a period not exceeding 6 months |
| Level 2 | Above 40% to 60% | Debarred from submitting a revised script for one year |
| Level 3 | Above 60% | Registration for that programme shall be cancelled |
Repeat offences are punished one level higher than the previous one. Publications carry a separate schedule under 12.2, adding withdrawal of the manuscript, denial of annual increments and bars on supervising new research students.
Penalties are not automatic
Section 12 states that penalties apply only after academic misconduct “has been established without doubt, when all avenues of appeal have been exhausted and individual in question has been provided enough opportunity to defend himself or herself in a fair or transparent manner.” A percentage on a report is not, by itself, a finding.
What does not count toward the percentage
Section 7 lists the exclusions, and applying them properly moves more theses between bands than any rewriting does.
- All quoted work reproduced with the necessary permission and attribution.
- All references, bibliography, table of contents, preface and acknowledgements.
- All generic terms, laws, standard symbols and standard equations.
- Common knowledge or coincidental terms, up to fourteen consecutive words.
That last one is specific and widely unknown. A run of fourteen identical words is not, on its own, within the definition the regulation applies — which matters in methods sections where standard procedures are described in standard language.
The same note requires that the original work — abstract, summary, hypothesis, observations, results, conclusions and recommendations — carry no similarities at all. Boilerplate is forgiven; your findings are not.
These rules may not apply to your document
The definition of “Script” in section 2(o) is narrower than most people assume. It covers research papers, theses, dissertations, book chapters and books submitted for assessment or publication at master's and research level.
It then states explicitly that this “shall exclude assignments / term papers / project reports / course work / essays and answer scripts”.
Coursework is outside this regulation
If you are an undergraduate worrying about a similarity score on an essay, the levels above are not the rules that govern it. Your institution will have its own policy, which may be stricter or more lenient. Quoting UGC bands at a coursework panel is quoting the wrong document.
There is no UGC threshold for AI
This is the part that matters most right now, and it is simply a matter of reading the text. The 2018 regulations were notified before generative writing tools were in general use, and they contain no mention of artificial intelligence, AI-generated content, or machine-generated text anywhere.
Everything the regulation defines — the four levels, the penalties, the exclusions, the panels — is attached to similarity: text matching an existing source. An AI detection score measures something different, and no nationally notified threshold for it exists.
| Similarity score | AI score | |
|---|---|---|
| Defined in UGC 2018 | Yes, with four bands | Not mentioned |
| Penalties attached | Yes, section 12 | None specified |
| Exclusions defined | Yes, section 7 | Not applicable |
| Measures | Matching existing text | Statistical resemblance to generated text |
Two things follow, and they point in opposite directions. A student panicking over a 30% AI score is worrying about a number with no defined regulatory consequence. But that is not permission: institutions set their own AI policies, undisclosed AI use is treated seriously under most of them, and the absence of a national threshold means your department's rule is the one that governs you.
If an AI score has been raised against your thesis, the practical ground is different from the regulatory one — see what the false-positive evidence actually shows and working from a DrillBit report.
Who decides, and in what order
The regulation builds a two-tier process rather than leaving it to a single marker.
- Anyone with appropriate proof may report a suspected case to the Departmental Academic Integrity Panel.
- The DAIP investigates at department level.
- The Institutional Academic Integrity Panel imposes any penalty, considering severity.
- Appeals must be exhausted before a penalty applies, and you must have had a fair opportunity to defend yourself.
Practically, that means asking for the full report with the highlighted passages rather than the summary percentage, and putting your account in writing early. The same evidence that resolves an AI allegation — drafts, version history, supervision notes — works here.
Want the method, not just the tool?
Turnitin vs DrillBit: Which Does Your University Use? — Most advice about AI reports is written about Turnitin. Across Indian higher education many universities run DrillBit — and the reports, the access and the confidence levels are not the same.
Try it on your own document
Write, cite, humanize and check your AI score in one editor. The free plan includes limited AI chat and AI detection each day, plus three humanizer runs — no card needed.
Open the editor →