A recent study published in Nature has drawn significant attention: 13 major large language models (LLMs) were shown to produce fabricated academic content under certain prompts—including fictional studies, invented datasets, and narratives that support pseudoscience.

Source: Gibney, E. (2026). Hey ChatGPT, write me a fictional paper: these LLMs are willing to commit academic fraud. Nature, 03 March. https://doi-org.libezproxy.um.edu.mo/10.1038/d41586-026-00595-9

Generative AI has become a common learning aid for university students, but this study reminds us that an AI system being “helpful” does not mean it is accurate and may even introduce academic risks. To help students enhance their AI literacy, the University of Macau Library has compiled key insights and practical recommendations to support your safe and responsible use of AI tools.

I. What Risks Can AI Introduce in Academic Research?

Researchers prompted 13 AI tools with five categories of questions, ranging from general queries to explicit requests to fabricate academic content. Their findings include:

• Most AI tools generate fake paper and fabricated data

Some versions of GPT, Grok, and other models produced seemingly plausible research drafts—complete with made‑up datasets, charts, and results.

• Only a few AI tools refused inappropriate requests

A small number of models (e.g., some versions of Claude) adhered to ethical principles more consistently, but none were perfectly safe.

Worryingly, even when an AI refused to fabricate data, it sometimes offered instructions on how to commit academic fraud, enabling users to perform the misconduct themselves.

II. Why Should These Findings Concern Us?

Irresponsible AI usage may harm the academic community in several ways:

1. A surge of low‑quality or fabricated papers could overshadow meaningful scientific work.
2. False data may contaminate entire research fields, affecting future studies and meta‑analyses.
3. Academic trust systems may be undermined, worsening ethical risks under the “publish or perish” pressure.
4. Faculty, researchers, and reviewers may face heavier workloads verifying authenticity.

AI is not intentionally deceptive—its primary function is to generate seemly plausible content, not to verify truth.

III. How Can You Avoid AI‑Related Academic Pitfalls?

The Library Recommends Three Key Practices

1. Treat AI with Healthy Skepticism:

Think of it as a “talkative but not fully reliable study partner.”

AI may express incorrect information in fluent, confident language. Always ask:
“Where can I verify what the AI just told me?”

Not merely:
“Is the AI correct?”

           ———————

2.Verify Information Using Library Databases (Highly Recommended)

2. Library databases and academic resources provide peer‑reviewed, authoritative information. They will help you:

   • Check whether AI‑generated references or studies actually exist
   • Locate credible primary sources and authoritative research
   • Validate theories, models, and research methodologies

   UM Library offers access to more than 260 databases.
   You can explore them via:
   Library Homepage → BoQuest → Databases
   UM Pass allows access from both on- and off‑campus (Note: VPN is not permitted).

   Commonly used databases @UM include:

   • ScienceDirect AI / Leap Space
   • Web of Science
   • Scopus
   • IEEE Xplore
   • Springer Nature Link
   • Wiley
   • Taylor & Francis
   • PubMed
   • JSTOR
   • SAGE
   • ProQuest
   • EBSCO
   • and so on.

   Library Website
   https://library.um.edu.mo/

           ———————

3. Strengthen Your Work with Your Own Research Efforts

Authentic academic research relies on genuine materials, such as:

• • Experimental results
  • Surveys and interviews
  • Field observations
  • Other primary data
  • Critical reading and analysis

AI can help with phrasing and summarization, but it cannot replace your inquiry, validation, or reasoning.

IV. Conclusion: AI Is a Tool—The Library Is Your Academic Support

When using AI, remember this three‑step workflow:

① AI Assistance → ② Library Verification → ③ Your Own Evidence & Data

If you would like to:

• Improve your AI literacy and avoid unintentional academic risks
• Strengthen your research and information‑seeking skills
• Learn citation management and plagiarism avoidance

You are welcome to join our Library workshops or consult our Reference Librarians:

Tel: (853) 8822 8170
Email: lib_ref@um.edu.mo

The library will remain your most reliable academic partner in the age of AI.

Final Thoughts

AI can help you write, but it cannot help you think.
The Library can help you search, but it cannot help you skip the hard work.

What truly carries you further in your university journey, and beyond,
is your curiosity, critical thinking, and desire to learn.

Don’t let AI lead you astray.
Let the Library bring you back to the world of genuine knowledge.

Library Website
https://library.um.edu.mo/

UM Library WeChat Service Account