UGC AI Policy for PhD Thesis 2026: What Every Indian Researcher Must Know

UGC AI Policy for PhD Thesis 2026: What Every Indian Researcher Must Know

If you’ve used ChatGPT, Grammarly, or any AI tool while working on your PhD thesis, you’ve probably asked this question: what does UGC actually say about it? The answer isn’t a single rule or a clean percentage. The UGC AI policy for PhD thesis 2026 sits across three regulatory layers — and the AI detection process is entirely separate from the similarity check most researchers already know. Here’s what actually triggers a flag, how the two scores work differently, what examiners look at, and how your institution type shapes your obligations.

• What is UGC’s AI Policy for PhD Theses?
• How Does UGC AI Detection Work?
• Why UGC’s AI Policy Matters for Indian Researchers
• Common Misconceptions About UGC’s AI Rules
• What to Do If Your Thesis Is Flagged for AI Content
• Conclusion

What is UGC’s AI Policy for PhD Theses?

India has no standalone UGC AI policy document. No separate circular, no AI-specific threshold, no dedicated regulation published for PhD theses as of 2026. What exists instead is a three-layer regulatory framework that treats AI-generated content under academic integrity rules already in force.

The foundation is the UGC (Promotion of Academic Integrity and Prevention of Plagiarism) Regulations, 2018. These regulations classify any unacknowledged content — whether copied from a journal, a website, or generated by an AI tool — as plagiarism. Submitting AI-generated text without disclosure is a direct violation. Not a grey area.

Alongside this, the 2020 PRAGYATA guidelines address AI and digital tools in Indian education broadly, recommending disclosure and responsible use. The third layer is your institution’s own PhD ordinance — which is why a researcher at Panjab University and one at JNU may face quite different rules on the same submission. Check your university’s ordinance before submitting.

In practice, the framework permits AI for supporting tasks. What it prohibits is presenting AI-generated writing as your own original scholarship. The core distinction isn’t whether you used AI — it’s whether you disclosed it honestly and whether the intellectual contribution is genuinely yours.

• Permitted with disclosure: Grammar correction, language improvement, formatting, outlining chapters, minor editing support
• Prohibited: Submitting AI-generated text as original writing, using AI-fabricated references or data results, concealing AI use while declaring none

How Does UGC AI Detection Work?

Most researchers understand similarity scores. AI detection is a completely separate process — and confusing the two is one of the most damaging assumptions to carry into your thesis submission.

A similarity score works through text-matching. The software compares your thesis against a database of academic papers, websites, and books, flagging passages that appear in other sources. It measures how much of your text has appeared elsewhere. High similarity usually reflects poor paraphrasing, missing citations, or genuinely copied content.

An AI content score works on an entirely different mechanism: perplexity analysis. AI-generated text is “perplexity-flat” — its word choices follow consistently probable statistical patterns that human writing doesn’t. When DrillBit-Extreme or Turnitin’s AI detector analyses your text, it estimates what percentage of the writing shows this flat, predictable pattern and returns it as an AI-generated percentage.

Here’s the critical point: these are two separate reports with independent results. Your thesis can have a low similarity score (no copied text detected) while registering a high AI score (prose that reads as statistically predictable). A thesis with many properly cited quotations may show higher similarity but a near-zero AI score. Clearing one does not clear the other.

Since October 2023, DrillBit-Extreme has been integrated into India’s national ShodhShuddhi portal (INFLIBNET). Every PhD thesis submitted through the national system now automatically generates both reports. Your institution doesn’t need to separately request the AI check — it runs by default.

One caveat worth knowing: AI detection tools are approximately 60–75% accurate even after manual paraphrasing. False positives are real. Non-native English speakers writing in a formal academic register — consistent grammar, technical vocabulary, disciplined sentence structure — can produce perplexity-flat prose that looks AI-generated but isn’t. Few supervisors mention this upfront. This is precisely why no institution can penalise you on an AI score alone. Human review alongside the report is required.

Why UGC’s AI Policy Matters for Indian Researchers

The consequences of a flagged thesis follow a four-tier penalty system — and most researchers learn about the tiers after submission.

The official UGC levels, as defined in the 2018 Regulations, apply to similarity scores. Institutions are increasingly applying the same scrutiny to combined AI + similarity findings:

• Level 0 (0–10% similarity): Accepted. No action required.
• Level A (10–40% similarity): Revise and resubmit within six months. Evaluation paused until resubmission is cleared.
• Level B (40–60% similarity): One-year debarment from resubmission. A major delay to your PhD timeline.
• Level C (above 60% similarity): PhD registration cancellation — applied after institutional review committee deliberation.

No equivalent official AI percentage threshold exists in UGC Regulations as of 2026. In practice, institutions flag submissions where the AI score exceeds 15–30%, with 20% being the most commonly cited benchmark. But this is not a regulatory floor — an AI score alone cannot automatically trigger Level A–C penalties. It triggers an investigation and human review.

Your institution type shapes your obligations significantly. Rules vary:

• Central universities, IITs, NAAC A+ institutions: Mandatory AI-use disclosure statements and methodology notes are typically required at the point of thesis submission.
• State universities: Apply the UGC 2018 similarity rules as a baseline; AI policy implementation varies widely. Check your specific university’s PhD ordinance before submitting.
• Deemed universities: Must follow the UGC framework as a minimum; many have added internal AI policies on top.
• AICTE-affiliated institutions: AICTE has separately mandated that AI-detection tools be used alongside similarity software for all thesis evaluations.
• IITs and IIMs: Many operate stricter internal policies beyond the UGC minimum — some require chapter-level AI-use declarations in the methodology section.

For a fuller breakdown of how the similarity penalty tiers are applied at Indian universities, see our guide on UGC AI plagiarism guidelines 2026.

Common Misconceptions About UGC’s AI Rules

Misinformation about AI in PhD research spreads fast through department corridors and WhatsApp groups. Four myths are worth correcting before you finalise your submission.

Myth 1: “There’s a safe AI percentage, just like the 10% similarity threshold.”

There isn’t. The UGC Regulations 2018 set four official similarity levels with exact penalties. No equivalent AI percentage threshold exists in any UGC circular or notification as of 2026. Institutions set their own informal benchmarks — typically 15–30% — but none of these carry the same regulatory standing as the Level 0–C similarity tiers. An examiner cannot automatically trigger a penalty based on an AI score alone.

Myth 2: “Paraphrasing will fool the AI detector.”

This stopped being reliable in 2024. AI detection tools now maintain approximately 60–75% accuracy even after manual paraphrasing. The detectors analyse statistical writing patterns — not specific phrases — so rewriting individual sentences without changing your underlying prose style leaves the perplexity signature largely intact. Heavy paraphrasing reduces the AI score but rarely eliminates it. In practice, passages that survive paraphrasing still read like lightly restyled AI output — enough for a trained examiner to notice.

Myth 3: “A low similarity score means I’ve passed both checks.”

Not true — these are separate reports with separate results. A thesis can clear the similarity check (low textual overlap with existing sources) while still registering a high AI score (writing that reads as statistically predictable). Both reports are generated automatically through ShodhShuddhi now. Clearing one does not clear the other.

Myth 4: “All Indian universities follow the same AI rules.”

They don’t. Central universities and IITs typically require formal AI-use disclosures; many state universities haven’t issued any formal AI-specific policy and rely on the 2018 plagiarism framework only. If your university hasn’t published AI-specific rules, disclosing AI use proactively in your thesis methodology section remains the safest approach regardless of what the regulations technically require.

What to Do If Your Thesis Is Flagged for AI Content

A flagged AI score isn’t automatically a disciplinary matter. Here’s what to do, step by step.

1. Get the detailed report, not just the headline number. Request the full DrillBit or Turnitin AI report, which shows exactly which sections scored high. A 25% overall AI score concentrated in your literature review is a different situation from the same score spread across your original research chapters.
2. Compile evidence of authorship. Gather your research notes, draft versions, reading lists, fieldwork records, and supervisor correspondence that document the development of your ideas. If an examiner questions authorship, this trail of evidence is your strongest defence — a score is not conclusive proof of misconduct.
3. Identify the root cause honestly. Was it a chapter drafted with significant AI assistance? Formal academic writing that triggers false-positive perplexity-flat scores because you write in a non-native register? Or a methodology section written in standard disciplinary template language? Each situation calls for a different response.
4. Rewrite flagged sections from scratch. If the AI involvement was real in specific passages, rewrite them using your notes and primary data. Focus on your own analytical voice — vary sentence length, include your specific observations, tie claims directly to your own research findings.
5. Consult your supervisor before responding formally. Your supervisor can advise on your institution’s internal process, help draft a response to the examination committee, and provide documented testimony about the development of your work.
6. Know your appeal rights. The UGC Regulations provide for institutional review processes. If you believe a finding is a false positive, you have the right to contest it with supporting evidence through your institution’s academic integrity committee.

To understand how the DrillBit AI detection report is structured and what each section means, see our complete guide to UGC AI plagiarism guidelines for Indian PhD students.

Conclusion

The UGC AI policy for PhD thesis 2026 isn’t a single rule or a percentage cap. It’s a layered framework built on the 2018 Plagiarism Regulations, the 2020 PRAGYATA guidelines, and your own institution’s ordinances — and the way these three interact in practice varies more than most researchers realise until they’re already at the submission stage. Similarity penalty tiers are official and fixed. AI score thresholds are informal and institution-specific. Across every type of Indian university, though, the underlying principle holds: your research must be yours, AI assistance must be disclosed, and you must be able to defend your intellectual contribution.

• No national AI% cap: UGC has not set an official AI content threshold — only the similarity levels (0–10% / 10–40% / 40–60% / above 60%) carry regulatory weight under the 2018 Regulations.
• Two separate reports: Similarity and AI scores are generated independently through ShodhShuddhi — a low similarity score doesn’t guarantee a low AI score.
• Institution type shapes your obligations: Central universities and IITs require formal disclosure; state university rules vary — check your specific PhD ordinance.
• Disclose and document: Any AI assistance used for grammar, formatting, or outlining should be declared in your methodology section. Proactive disclosure protects you if the AI score is later queried.