⇣ ⇣ MASON FACULTY: See virtual office and Lunch & Learn details. ⇣ ⇣

Innovation Insight banner

August 8, 2023 | View as Webpage | academicinnovation@mason.wm.edu

Citing Generative AI

As use of ChatGPT proliferated in the wake of its November 2022 release, many faculty and students began to ask, “Should we use it? And if we use it, how should we cite it?” In this issue of Innovation Insight, we will discuss six key questions about this process, exploring issues of authorship, credibility, and intellectual property. 

An Interview with Anna Milholland,

Head of McLeod Business Library

This is such a new field; does generative AI change how we cite sources?

Citation still serves the same purposes: crediting creators and continuing scholarly conversation. 

The U.S. Copyright Office defines authorship as a right exclusive to humans. Therefore, the U.S. Copyright Office and many scholarly journals and societies do not currently consider generative AI tools to be authors under U.S. law. So, works using them may not be protected by copyright. However, you must still acknowledge the tool’s contributions. Many professional associations, including the American Psychological Association, the Modern Language Association, and University of Chicago Press, have issued guidance on how to do this.

What are current best practices?

While guidelines will vary by discipline, general best practices for citing generative AI using APA style include:

  • Acknowledging and describing the use of the generative AI tool, including a description of the tool, prompt engineering process, and how it is used (e.g., drafting, summarizing, analyzing data).
  • Attributing authorship to the company (i.e., Open AI, Anthropic, Google) not the generative AI software. Remember, software is not considered an author or creator, but the developing company is. 
  • Including the version and access date in the citation, as different versions may provide varied outputs, and this information helps maintain a scholarly record. 

Should I consider requiring students to submit a list of all the prompts and all responses from generative AI?

The decision to request a complete prompt history depends on what you value in the assignment. If understanding a student's thought or research process is important, then it could be useful to ask for all prompts. Students using generative AI should already acknowledge and describe its usage in their work. Requesting all prompts and responses may lead to a longer document, but the necessity and length will depend on your specific requirements and the nature of the assignment.

Should I prompt generative AI for references? What happens if I do and what happens if I don’t?

You can ask for references from generative AI, but be aware that these tools may produce incorrect information or citations. It's not advisable to rely on generative AI for research on ill-defined, obscure, non-mainstream, non-Western, personal, or unfamiliar topics without verifying the information from multiple sources. Asking for references allows you to cross-reference the information for accuracy and relevance, but don't expect the AI to consistently provide accurate or comprehensive reference lists or bibliographies. 

If a student uses generative AI to produce work and that work contains information that is not properly cited (i.e., plagiarized), is the student responsible?

Large language models like ChatGPT can produce text that is plagiarized, false, or illegal, because they do not understand the meaning, context, or implications of the words they generate. Anyone who reproduces someone else’s idea without proper citation commits plagiarism, whether it is done knowingly or unknowingly. To avoid plagiarism and avoid promoting misinformation, students should cross-reference statements from generative AI with multiple sources and cite an original scholarly source in addition to acknowledging the use of the AI tool and providing prompts if requested. Instructors can seek assistance from the McLeod Business Library to help students accomplish this goal. 

Should I consider using AI-detection tools (e.g., Turnitin, GPTZero) to detect generative AI?

Tools exist that claim to differentiate AI-produced content from human-written work, but they often generate false results. A study (not yet peer-reviewed) submitted to the International Journal for Educational Integrity found these tools to be neither accurate nor reliable, often misclassifying AI-generated text as human-written. If you decide to use these AI-detection tools, the McLeod Business Library recommends that you do so with care.


Mason Faculty

Virtual Office Hours

Thursdays, July 20 - August 24, 2023

12:00 Noon

Zoom link

Hosted by Academic Innovation, McLeod Business Library and Center for Online Learning. Browse to tinyurl.com/masonmeet/ and drop in Zoom to join the discussion.

Lunch & Learn

Friday, September 22

12:30 pm - 1:30 pm

Miller 1018

Zoom link

Professors Phil Wagner, Dawn Edmiston, and Jim Bradley will share how they are using generative AI in business education. Join the discussion in-person or remotely at tinyurl.com/masonlunchlearn/.


The following are examples that were modified from the APA Style Blog's article, "How to Cite ChatGPT" (April 2023) for acknowledging and citing text-based generative AI outputs:

Acknowledgement examples:

In the introduction section: “The author acknowledges the use of ChatGPT-4 (Mar 14 version) to draft portions of this article by prompting GPT-4 to….”

In the Methods section: “The author acknowledges the use of ChatGPT-4 (Mar 14 version) to conduct a thematic analysis by coding qualitative data…”

Parenthetical citation: (Open AI, 2023)

Narrative citation: OpenAI (2023)

Direct Quotes: When prompted with “Is the left brain right brain divide real or a metaphor?” the ChatGPT-generated text indicated that although the two brain hemispheres are somewhat specialized, “the notation that people can be characterized as ‘left-brained’ or ‘right-brained’ is considered to be an oversimplification and a popular myth” (OpenAI, 2023).

Reference List: OpenAI. (2023). ChatGPT3.5 (Mar 14 version) [Large language model]. Accessed July 10, 2023, from https://chat.openai.com/chat.

Note: We recommend that direct quotes be used sparingly and, when at all possible, replaced with a direct quote from an original (non-generative AI) source.

LIBRARY CORNER with Anna Milholland, McLeod Library

Questions? Ask a Librarian. If you have questions about generative AI and citations, authorship, copyright, and scholarly publishing, please contact Anna Milholland, Head of McLeod Business Library, at anna.milholland@mason.wm.edu or submit your question to https://tinyurl.com/AskAboutAI.


Authorship and copyright: Can AI be an inventor on a patent or author of a copyrighted work?

Join Nathaniel Lipkus, Partner, Intellectual Property Group with Osler, Hoskin & Harcourt LLP, a leading business law firm advising Canadian and international clients from offices across Canada and in New York. He asks whether AI can be an inventor on a patent or author of a copyrighted work. 


How to cite ChatGPT (apa.org)

Learn about:

  • quoting or reproducing the text created by ChatGPT in your paper and
  • referencing ChatGPT or other AI models and software

When AI Chatbots Hallucinate (nytimes.com)

ChatGPT is not alone in fabricating information.

ChatGPT and Fake Citations (duke.edu)

Learn what ChatGPT is good for and what it is not good for.

Testing of Detection Tools for AI-Generated Text (arxiv.org)

This paper digs deeper to examine the general functionality of detection tools for artificial intelligence generated text and evaluate them based on accuracy and error type analysis.

Alan B. Miller Hall 3046 | academicinnovation@mason.wm.edu