Using AI tools like ChatGPT in your PhD

Is using ChatGPT allowed in a PhD?

It depends on what you use it for, and on your university. There is no single national rule in India that says yes or no, and there is a large difference between asking a tool to explain a regression output and asking it to write your literature review.

The principle that every publisher and most universities share is simple: you are responsible for every word, number and reference in your work, and you must be able to explain and defend all of it. A thesis is assessed as evidence of your ability to do research. Text you did not write is not evidence of that, however good it is. That is also where our own ethics policy draws the line for the support we offer.

The table is our reading of current practice, not a rule. Some universities are stricter, for example requiring disclosure even of grammar tools, and a few allow more. When the university has issued a circular, it overrides anything here.

What do journals and publishers require?

If you plan to publish from your thesis, the journal’s policy matters as much as the university’s. The main positions, as checked on the publishers’ own pages in September 2026:

Three things are common to all of them. AI cannot be an author. The human authors are accountable for accuracy, including any errors the tool introduced. And fabricated data, results or references are misconduct whether a person or a tool produced them.

They differ on the details: where the disclosure goes (a separate statement, the Methods section, the acknowledgments) and what counts as minor editing. Read the specific journal’s author guidelines before you submit, because many journals add their own rules on top of the publisher’s.

What a disclosure looks like

Short and specific. For example: “During the preparation of this work the author used ChatGPT (OpenAI, [model and version]) to improve the readability of the introduction and discussion. The author reviewed and edited all output and takes full responsibility for the content.” Name the actual tool and version you used, and do not describe more or less use than there was.

What do UGC and Indian universities say about AI?

The UGC (Promotion of Academic Integrity and Prevention of Plagiarism in Higher Educational Institutions) Regulations, 2018 are the national rules on plagiarism in theses. They were written before generative AI was in common use and do not mention it. At the time of this review we could not find a UGC regulation or notice on ugc.gov.in specifically governing generative AI in PhD theses. You may see websites claiming “new UGC AI guidelines” with detailed rules; unless they link to an official UGC document, treat them with caution.

In practice, universities apply the 2018 regulations and their own academic integrity policies: presenting text you did not write as your own is treated as an integrity problem whatever produced it. Several Indian universities and institutes have issued their own circulars or policies on AI use, and more are expected. Ask your research cell or supervisor whether your university has one, and keep a copy.

The similarity levels and penalties under the 2018 regulations are explained in our plagiarism limits guide.

Can Turnitin or DrillBit detect AI writing?

Both Turnitin and DrillBit, the checker INFLIBNET provides to universities through ShodhShuddhi, now offer AI-writing indicators alongside similarity reports. What they give is a probability estimate, not proof.

Turnitin’s own guidance is careful about this. It states that institutions should not penalise a student on the AI writing score by itself, and that the result needs human judgement alongside the institution’s policy. Because Turnitin found a higher incidence of false positives at low scores, its report shows an asterisk instead of a percentage for AI scores between 1% and 19%. In August 2025 Turnitin announced that its detection also covers text likely modified by “AI bypasser” or humaniser tools, trained and tested for English.

What this means for you:

• False positives happen. Formulaic academic prose, heavily edited text and writing by non-native speakers of English can be flagged even when the scholar wrote it.
• Paraphrasing tools can raise the AI score. Text run through a rewriter to lower similarity may be flagged as AI-modified. Our paraphrasing guide explains why these tools cause trouble.
• A low AI score is not permission. Detectors miss things, and examiners and viva panels test understanding directly.

How to read the similarity side of the report is covered in the Turnitin report guide.

Why do AI tools invent references?

Chatbots generate text that looks like what usually follows a prompt. A reference list usually contains author names, a plausible title, a journal and a year, so the tool produces exactly that, whether or not the paper exists. The result can have real author names, a real journal and a DOI that leads nowhere or to a different paper.

A thesis with even one fabricated reference is a serious problem. Examiners do check, and a non-existent citation suggests that the scholar did not read what they cited. Elsevier’s and Springer Nature’s policies both name fabricated citations as unacceptable.

• Never copy a reference from a chatbot into your reference manager. Find the paper in Google Scholar, Scopus, Web of Science or the publisher’s site, and import it from there.
• Tools that search real databases and link to sources reduce this risk but do not remove it. Open the source and check that it says what the tool claims.
• Keep references in Zotero or Mendeley, imported from the database or DOI, so each one is traceable.

How do you protect yourself if your work is questioned?

The best protection is evidence of how your work was produced. Start keeping it now, not after a report flags something.

• Keep a simple AI-use log: date, tool and version, what you asked, what you used, and how you checked it.
• Keep your own drafts and version history (Word version history, Google Docs history, or dated files).
• Ask your supervisor, in writing, which uses they are comfortable with, and follow any university circular.
• Verify every factual statement, number and reference that a tool gives you against the original source.
• Never paste participant data, unpublished results or a manuscript under review into a public AI tool.
• Disclose AI use in the form your university or target journal asks for.

Confidentiality and data protection

Interview transcripts, survey responses with identifiers, patient records and unpublished results should not go into a public AI tool. Your consent forms and ethics approval promised participants how their data would be handled, and a public chatbot was almost certainly not part of that promise. India’s Digital Personal Data Protection Act, 2023 also applies to personal data. If you want to use an AI tool on research data, discuss it with your supervisor and ethics committee first, and prefer tools your institution has approved.

If you are accused

Stay calm and ask for the specific evidence. Share your drafts, version history, notes and AI-use log. Offer to explain any part of the work in person, which is the strongest evidence you can give. You can also point, politely, to Turnitin’s own guidance that a detector score needs human judgement alongside it. A scholar who can show how the work developed is in a strong position.