ChatGPT generated unverified academic references despite explicit instructions

I subscribed to ChatGPT Go specifically for thesis and academic research.

My very first prompt explicitly required only verified references from trusted academic sources such as NCERT, Shodhganga, UGC-CARE journals, INFLIBNET, ICSSR, NIEPA, and Indian university theses. I also clearly instructed ChatGPT not to generate fabricated or unverifiable citations.

Despite these instructions, ChatGPT confidently generated author names and study titles that I could not verify. As a research student, this cost me significant time because I had to manually check every citation.

For academic users, this is a serious issue. If ChatGPT cannot verify a reference, it should clearly say so instead of presenting it as a real study.

I am sharing the conversation and a screenshot so the issue can be reproduced.

Conversation:

Screenshot attached below.

I hope OpenAI improves citation verification for academic research. Accuracy is more important than confident-looking but unverified references.

This issue is reproducible using the shared conversation above.

Welcome the community @Rajnikant_Parmar,

If I have understood and had it translated correctly (please correct me if I’m misunderstanding), but this appears to be a very broad request: ten fully verified studies from several different repositories, followed by a detailed 700–1,000-word review of each one. Breaking the work into smaller stages may produce more reliable results.

I experienced something similar while writing my master’s thesis, ChatGPT occasionally referred to articles that weren’t included in my study. I found it more reliable to work with one source at a time.

Here’s few tips on how I would do:

I would first ask ChatGPT to locate one study and provide its direct source link or DOI without writing the review. After independently verifying that it exists, upload or link the full article and ask ChatGPT to analyse only that document. Repeat this process for each study.

Deep research⁠ may also help because it can be restricted to specified websites and provides source links. Nevertheless, academic references should always be checked manually, as ChatGPT can produce incorrect or fabricated citations⁠.

Hi @Rajnikant_Parmar,

As Larisa mentioned, ChatGPT can sometimes generate inaccurate or unverifiable citations, so it's always a good idea to independently verify references before using them in academic work.

For research tasks like this, you may also find Deep Research helpful, as it can provide source links and is designed for more in-depth research workflows. Even then, we recommend verifying any citations or references against the original sources before including them in your work.

Thanks for taking the time to share your experience.

~Smith