Plagiarism and AI Detection Limits for IEEE Papers: Why M.Tech Papers Get Desk-Rejected | Anushram

By Dr. Rajesh Kumar ModiPosted 21 November 2026

Get Started Today

Fill out the form and our experts will reach out to you

Plagiarism and AI Detection Limits for IEEE Papers: Why M.Tech Papers Get Desk-Rejected | Anushram
7 min read

Why good M.Tech papers get desk-rejected: similarity limits, UGC plagiarism levels, IEEE's AI disclosure rule and how to fix your score before you submit.

You worked for months. The results are good. Then the email arrives: "Desk rejected: similarity score too high." No reviewer even opened your paper.

This happens to many M.Tech students, and it's almost always avoidable. In this Anushram guide you'll learn how plagiarism and AI detection limits work for IEEE papers, what the rules actually say, and how to fix your paper before you submit.

Quick answer: Keep overall similarity as low as you can, ideally under 10% (UGC's "no penalty" level) and always below your conference's stated limit. Avoid large overlap with any single source. If you used AI to generate any text, figures or code, IEEE requires you to disclose it in the Acknowledgments. Grammar-only help is outside that rule. Check every reference is real, then run your own similarity check before submitting.

An Anushram guide for M.Tech researchers.

Similarity, plagiarism and AI detection: not the same thing

These three words get mixed up all the time. Here's the difference:

TermWhat it meansWho measures it
Similarity score% of your text that matches existing text somewhereSoftware (iThenticate, Turnitin)
PlagiarismUsing someone's words or ideas without creditA human decision (editor, committee)
AI detection scoreSoftware's guess at how much text was AI-writtenSoftware (e.g. Turnitin AI detector)

A similarity score is not proof of plagiarism. Matched references, standard terms and your own properly cited prior work all add to it. But editors use it as a first filter, so a high number can stop your paper before review.

What the rules actually say

IEEE: every paper is screened

IEEE checks conference papers with CrossCheck (now Crossref Similarity Check), which compares manuscripts against a large database of published papers and over 6 billion web pages (IEEE NPSS). It runs on Turnitin's iThenticate software (Crossref).

IEEE does not publish one universal percentage, so each conference sets its own limit. Examples from real calls for papers:

  • One IEEE conference stated that for IEEE Xplore, overall similarity should not exceed 30% and similarity to any single source should not exceed 10% (IC_ASET).
  • ICTIEE 2027 desk-rejects papers with a Turnitin similarity above 12% or an AI-detection score above 25% (EasyChair CFP).

UGC (India): four plagiarism levels

The UGC Regulations 2018 set these levels for academic and research publications (CASRAI summary):

LevelSimilarityWhat happens (publications)
Level 0Up to 10%Minor, no penalty
Level 1Above 10% to 40%Manuscript must be withdrawn
Level 2Above 40% to 60%Withdrawal plus further penalties
Level 3Above 60%Withdrawal plus the most serious penalties

The same rules let checks exclude common knowledge or coincidental terms of up to 14 consecutive words (S.S. Rana).

IEEE's AI policy

  • IEEE authors may use AI-generated content only if it's disclosed in the Acknowledgments section (IEEE CAS).
  • The disclosure must name the AI system, the sections where it was used, and how much it was used (IEEE AESS).
  • Using AI for editing and grammar is outside that policy; disclosing it is recommended but not required (IEEE CAS).
  • Some IEEE events state that non-existent (fake) references lead to desk rejection (DESY event policy).

A word on AI detectors

AI detectors are not perfect. When Turnitin launched its detector, it found more false positives when less than 20% of a document was flagged, and it now marks those scores with an asterisk (Higher Ed Dive). Still, many conferences use these scores as a filter, so you need to stay well below their limit.

How your paper is screened before review

Before any reviewer reads your paper, it has to pass three quick checks. Failing any one of them ends the process.

How an IEEE conference screens a submission · 3 gates

The exact checks vary by conference; some skip the AI check, others add a scope check by the program chair.

How to reduce similarity the right way

Don't use "paraphrasing" or "spinner" tools to hide copied text. Reviewers spot the strange wording, and it doesn't fix the real problem. Instead:

  1. Read the similarity report. Find which sources match and which sections are highest (usually the introduction and literature review).
  2. Rewrite from understanding. Close the source, explain the idea in your own words, then cite it.
  3. Summarise, don't quote. One sentence per paper in the literature review, focused on what matters for your work.
  4. Use quotation marks for any exact definition you must keep, with a citation.
  5. Rewrite your own thesis text. Copying from your earlier report, thesis or paper counts as self-plagiarism if it was published.
  6. Describe methods freshly. Don't copy standard method descriptions from a paper; explain them with your own notation.
  7. Check again after editing, until you're comfortably below your conference's limit.

Using AI tools safely in your IEEE paper

Use of AIOK?What to do
Fixing grammar and spellingYesDisclosure recommended, not required
Improving clarity of a sentence you wroteYesReview every change; disclosure recommended
Generating paragraphs, figures or code for the paperOnly with disclosureName the tool, the sections and the extent in Acknowledgments
Writing the literature review from AI summariesRiskyRead the real papers; AI often invents or misstates sources
Adding references suggested by AIOnly if verifiedOpen every paper and confirm it exists and says what you claim
Uploading someone else's manuscript to a public AI toolNoBreaks confidentiality (IEEE forbids this for reviewers)

Example disclosure: "The authors used ChatGPT (OpenAI) to improve the language of Sections I and II. All content was reviewed and edited by the authors, who take full responsibility for it."

Remember: you are responsible for every word, figure and reference, whatever tool helped.

Pre-submission integrity checklist

  • ☐ Similarity report checked; overall score well below the conference limit.
  • ☐ No single source contributes a large block of matched text.
  • ☐ Every idea, figure and dataset from others is cited.
  • ☐ Text reused from your thesis or earlier paper is rewritten or cited.
  • ☐ Every reference opened and verified as real.
  • ☐ Any AI-generated content disclosed in Acknowledgments (tool, sections, extent).
  • ☐ AI-detection score checked, if your conference sets a limit.
  • ☐ Paper not submitted to any other conference or journal at the same time.
  • ☐ Guide has read and approved the final version.

Frequently asked questions

Q1. What is the acceptable plagiarism percentage for an IEEE conference paper? 
There's no single IEEE number; each conference sets its own. Aim for under 10% overall, with no large overlap from one source, and always stay under the limit in the call for papers.

Q2. Do references count in the similarity score? 
They can, unless the checker is set to exclude the bibliography. Editors usually look past matched references, but check your report.

Q3. Can I use ChatGPT for my IEEE paper? 
For grammar and editing, yes. If AI generated any content, IEEE requires you to disclose it in the Acknowledgments with the tool, sections and level of use.

Q4. My AI score is high but I wrote everything myself. What now? 
Detectors can be wrong, especially on formal, formulaic writing. Keep your drafts and notes as proof, vary sentence structure, and contact the program chair if you're flagged.

Q5. Is reusing my own M.Tech thesis text plagiarism? 
If the thesis is published or online, copying it without citation is self-plagiarism. Rewrite it and cite the thesis where needed.

Originality is your best shortcut

A desk rejection for similarity or AI use wastes weeks. A clean, original paper with honest disclosure goes straight to reviewers. Write in your own words, cite generously, verify every reference, and check your score before you submit.

Your next step: Run your current draft through your university's similarity checker today and fix the top three matching sources.

Need help preparing your M.Tech paper? Visit anushram.com for research guides and support.

Read next on Anushram: How to Write a 6-Page IEEE Paper That Gets Accepted · Top Upcoming IEEE Conferences in India 2027

  • #plagiarism limit for IEEE paper
  • #similarity index for IEEE conference
  • #IEEE AI policy
  • #Turnitin AI detection
  • #UGC plagiarism levels
  • #how to reduce plagiarism in research paper
  • #Anushram

Reader Reviews

Loading…
Your rating

Get Started Today

Fill out the form and our experts will reach out to you