Skip to Main Content

Artificial Intelligence (AI) in Academics

How to Cite AI Sources

Specific rules for citing AI tools are not yet available from the American Medical Association (AMA). However, some sections of the style guide have been updated to provide general guidance for authors. A Journal of the American Medical Association (JAMA) editorial also provides guidance for authors and peer reviewers. Highlights of these resources are outlined below. 

AMA Manual of Style, Updated Sections

Section 5.1.12

  • "Nonhuman artificial intelligence, language models, machine learning, or similar technologies cannot be listed as authors..."

  • "Authors should report the use of artificial intelligence, language models, machine learning, or similar technologies to create content or assist with writing or editing of manuscripts in the Acknowledgment section or Methods section if this is part of formal research design or methods."

Section 5.2.1.1

  • Authors are required to "include a description of the content that was created or edited and the name of the language model or tool, version and extension numbers, and manufacturer."

Section 14.5.2

  • "For AI tools, including large language models (LLMs), machine learning, generative AI such as chatbots, and related technologies, use a nonproprietary descriptive term. In research articles, at first mention in the abstract and in the text, provide the brand name in parentheses along with the version or extension number, manufacturer or owner, and date(s) used.

    • Example: "On June 12, 2023, the original full text of the question was put into a fresh chatbot session (ChatGPT, model GPT-4, OpenAI) and the generated responses were saved."

JAMA Editorial

  • "[JAMA] policies preclude the inclusion of nonhuman AI tools as authors and require the transparent reporting of use of such tools in preparing manuscripts and other content and when used in research submitted for publication.

  • "...submission and publication of clinical images created by AI tools is discouraged, unless part of formal research design or methods."

Authors are required to respond to the following question when submitting their work to JAMA Network journals.

  • Did you use AI, a language model, machine learning, or similar technologies to create or assist with creation or editing of any of the content in this submission (eg, text, tables, figures, video)?

If authors respond yes to the question, they are required to provide a description of the AI-generated content and confirm that they are responsible for the integrity of the content. 

  • "Please provide a description of the AI-generated content that is included in this submission and the name of the model or tool used, version and extension numbers, and manufacturer..."

  • "Please confirm that you take responsibility for the integrity of the content generated by these tools and that you have provided a description of such generated content and the name of the model or tool used, version and extension numbers, and manufacturer in the Acknowledgment or Methods section of the manuscript."

Source: https://mdanderson.libguides.com/cite/ai

Quoting or reproducing the text created by ChatGPT in your paper

If you’ve used ChatGPT or other AI tools in your research, describe how you used the tool in your Method section or in a comparable section of your paper. For literature reviews or other types of essays or response or reaction papers, you might describe how you used the tool in your introduction. In your text, provide the prompt you used and then any portion of the relevant text that was generated in response.

Unfortunately, the results of a ChatGPT “chat” are not retrievable by other readers, and although nonretrievable data or quotations in APA Style papers are usually cited as personal communications, with ChatGPT-generated text there is no person communicating. Quoting ChatGPT’s text from a chat session is therefore more like sharing an algorithm’s output; thus, credit the author of the algorithm with a reference list entry and the corresponding in-text citation.

Example:

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

OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat

You may also put the full text of long responses from ChatGPT in an appendix of your paper or in online supplemental materials, so readers have access to the exact text that was generated. It is particularly important to document the exact text created because ChatGPT will generate a unique response in each chat session, even if given the same prompt. If you create appendices or supplemental materials, remember that each should be called out at least once in the body of your APA Style paper.

Example:

When given a follow-up prompt of “What is a more accurate representation?” the ChatGPT-generated text indicated that “different brain regions work together to support various cognitive processes” and “the functional specialization of different regions can change in response to experience and environmental factors” (OpenAI, 2023; see Appendix A for the full transcript).

Reference

OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat

Creating a reference to ChatGPT or other AI models and software

The in-text citations and references above are adapted from the reference template for software in Section 10.10 of the Publication Manual (American Psychological Association, 2020, Chapter 10). Although here we focus on ChatGPT, because these guidelines are based on the software template, they can be adapted to note the use of other large language models (e.g., Bard), algorithms, and similar software.

The reference and in-text citations for ChatGPT are formatted as follows:

OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat

  • Parenthetical citation: (OpenAI, 2023)
  • Narrative citation: OpenAI (2023)

Let’s break that reference down and look at the four elements (author, date, title, and source):

Author: The author of the model is OpenAI.

Date: The date is the year of the version you used. Following the template in Section 10.10, you need to include only the year, not the exact date. The version number provides the specific date information a reader might need.

Title: The name of the model is “ChatGPT,” so that serves as the title and is italicized in your reference, as shown in the template. Although OpenAI labels unique iterations (i.e., ChatGPT-3, ChatGPT-4), they are using “ChatGPT” as the general name of the model, with updates identified with version numbers.

The version number is included after the title in parentheses. The format for the version number in ChatGPT references includes the date because that is how OpenAI is labeling the versions. Different large language models or software might use different version numbering; use the version number in the format the author or publisher provides, which may be a numbering system (e.g., Version 2.0) or other methods.

Bracketed text is used in references for additional descriptions when they are needed to help a reader understand what’s being cited. References for a number of common sources, such as journal articles and books, do not include bracketed descriptions, but things outside of the typical peer-reviewed system often do. In the case of a reference for ChatGPT, provide the descriptor “Large language model” in square brackets. OpenAI describes ChatGPT-4 as a “large multimodal model,” so that description may be provided instead if you are using ChatGPT-4. Later versions and software or models from other companies may need different descriptions, based on how the publishers describe the model. The goal of the bracketed text is to briefly describe the kind of model to your reader.

Source: When the publisher name and the author name are the same, do not repeat the publisher name in the source element of the reference, and move directly to the URL. This is the case for ChatGPT. The URL for ChatGPT is https://chat.openai.com/chat. For other models or products for which you may create a reference, use the URL that links as directly as possible to the source (i.e., the page where you can access the model, not the publisher’s homepage).

Other questions about citing ChatGPT

You may have noticed the confidence with which ChatGPT described the ideas of brain lateralization and how the brain operates, without citing any sources. I asked for a list of sources to support those claims and ChatGPT provided five references—four of which I was able to find online. The fifth does not seem to be a real article; the digital object identifier given for that reference belongs to a different article, and I was not able to find any article with the authors, date, title, and source details that ChatGPT provided. Authors using ChatGPT or similar AI tools for research should consider making this scrutiny of the primary sources a standard process. If the sources are real, accurate, and relevant, it may be better to read those original sources to learn from that research and paraphrase or quote from those articles, as applicable, than to use the model’s interpretation of them.

We’ve also received a number of other questions about ChatGPT. Should students be allowed to use it? What guidelines should instructors create for students using AI? Does using AI-generated text constitute plagiarism? Should authors who use ChatGPT credit ChatGPT or OpenAI in their byline? What are the copyright implications?

On these questions, researchers, editors, instructors, and others are actively debating and creating parameters and guidelines. Many of you have sent us feedback, and we encourage you to continue to do so in the comments below. We will also study the policies and procedures being established by instructors, publishers, and academic institutions, with a goal of creating guidelines that reflect the many real-world applications of AI-generated text.

For questions about manuscript byline credit, plagiarism, and related ChatGPT and AI topics, the APA Style team is seeking the recommendations of APA Journals editors. APA Style guidelines based on those recommendations will be posted on this blog and on the APA Style site later this year.

Update: APA Journals has published policies on the use of generative AI in scholarly materials.

We, the APA Style team humans, appreciate your patience as we navigate these unique challenges and new ways of thinking about how authors, researchers, and students learn, write, and work with new technologies.

Reference

American Psychological Association. (2020). Publication manual of the American Psychological Association (7th ed.). https://doi.org/10.1037/0000165-000

Source: https://apastyle.apa.org/blog/how-to-cite-chatgpt

Citation, Documentation of Sources

Q. How do you recommend citing content developed or generated by artificial intelligence, such as ChatGPT? Many scholarly publishers are requiring its identification though also requiring human authors to take responsibility for it and will not permit the AI to have “authorship.”

A. You do need to credit ChatGPT and similar tools whenever you use the text that they generate in your own work. But for most types of writing, you can simply acknowledge the AI tool in your text (e.g., “The following recipe for pizza dough was generated by ChatGPT”).

If you need a more formal citation—for example, for a student paper or for a research article—a numbered footnote or endnote might look like this:

1. Text generated by ChatGPT, OpenAI, March 7, 2023, https://chat.openai.com/chat.

ChatGPT stands in as “author” of the content, and OpenAI (the company that developed ChatGPT) is the publisher or sponsor, followed by the date the text was generated. After that, the URL tells us where the ChatGPT tool may be found, but because readers can’t necessarily get to the cited content (see below), that URL isn’t an essential element of the citation.

If the prompt hasn’t been included in the text, it can be included in the note:

1. ChatGPT, response to “Explain how to make pizza dough from common household ingredients,” OpenAI, March 7, 2023.

If you’ve edited the AI-generated text, you should say so in the text or at the end of the note (e.g., “edited for style and content”). But you don’t need to say, for example, that you’ve applied smart quotes or adjusted the font; changes like those can be imposed silently (see CMOS 13.7 and 13.8).

If you’re using author-date instead of notes, any information not in the text would be placed in a parenthetical text reference. For example, “(ChatGPT, March 7, 2023).”

But don’t cite ChatGPT in a bibliography or reference list unless you provide a publicly available link (e.g., via a browser extension like ShareGPT or A.I. Archives). Though OpenAI assigns unique URLs to conversations generated from your prompts, those can’t be used by others to access the same content (they require your login credentials), making a ChatGPT conversation like an email, phone, or text conversation—or any other type of personal communication (see CMOS 14.214 and 15.53).

To sum things up, you must credit ChatGPT when you reproduce its words within your own work, but unless you include a publicly available URL, that information should be put in the text or in a note—not in a bibliography or reference list. Other AI-generated text can be cited similarly. Check back with us for updates on this evolving topic.

For some considerations on the use of AI in scholarly publishing and the responsibilities of authors, start with this position statement on authorship and AI tools from COPE (Committee on Publication Ethics).

[Editor’s update: Original answer updated to put the publisher before the date and to acknowledge tools like ShareGPT and A.I. Archives.]

[This answer relies on the 17th edition of CMOS (2017) unless otherwise noted.]

Source: https://www.chicagomanualofstyle.org/qanda/data/faq/topics/Documentation/faq0422.html

The MLA’s method for citing sources uses a template of core elements—standardized criteria that writers can use to evaluate sources and create works-cited-list entries based on that evaluation. That new technologies like ChatGPT emerge is a key reason why the MLA has adopted this approach to citation—to give writers flexibility to apply the style when they encounter new types of sources. In what follows, we offer recommendations for citing generative AI, defined as a tool that “can analyze or summarize content from a huge set of information, including web pages, books and other writing available on the internet, and use that data to create original new content” (Weed). 

You should

  • cite a generative AI tool whenever you paraphrase, quote, or incorporate into your own work any content (whether text, image, data, or other) that was created by it 
  • acknowledge all functional uses of the tool (like editing your prose or translating words) in a note, your text, or another suitable location 
  • take care to vet the secondary sources it cites (see example 5 below for more details)

See below for specific examples. And keep in mind: the MLA template of core elements is meant to provide flexibility in citation. So if you find a rationale to modify these recommendations in your own citations, we encourage you to do so. We’ve opened this post up for commenting, so let us know what you think and how you’re using and citing generative AI tools!

Using the MLA Template

Author

We do not recommend treating the AI tool as an author. This recommendation follows the policies developed by various publishers, including the MLA’s journal PMLA. 

Title of Source

Describe what was generated by the AI tool. This may involve including information about the prompt in the Title of Source element if you have not done so in the text. 

Title of Container

Use the Title of Container element to name the AI tool (e.g., ChatGPT).

Version

Name the version of the AI tool as specifically as possible. For example, the examples in this post were developed using ChatGPT 3.5, which assigns a specific date to the version, so the Version element shows this version date.

Publisher

Name the company that made the tool.

Date

Give the date the content was generated.

Location

Give the general URL for the tool.1

Example 1: Paraphrasing Text

Passage in Source

Screenshot of ChatGPT response about symbolism of the green light in The Great Gatsby.

Paraphrased in Your Prose

While the green light in The Great Gatsby might be said to chiefly symbolize four main things: optimism, the unattainability of the American dream, greed, and covetousness (“Describe the symbolism”), arguably the most important—the one that ties all four themes together—is greed.

Works-Cited-List Entry

“Describe the symbolism of the green light in the book The Great Gatsby by F. Scott Fitzgerald” prompt. ChatGPT, 13 Feb. version, OpenAI, 8 Mar. 2023, chat.openai.com/chat.

Example 2: Quoting Text

Passage in Source

Screenshot of ChatGPT reply to prompt about describing the symbolism of the green light in The Great Gatsby

Screenshot of ChatGPT reply to prompt about providing scholarly sources used to generate a description.

Quoted in Your Prose

When asked to describe the symbolism of the green light in The Great Gatsby, ChatGPT provided a summary about optimism, the unattainability of the American dream, greed, and covetousness. However, when further prompted to cite the source on which that summary was based, it noted that it lacked “the ability to conduct research or cite sources independently” but that it could “provide a list of scholarly sources related to the symbolism of the green light in The Great Gatsby” (“In 200 words”).

Works-Cited-List Entry

“In 200 words, describe the symbolism of the green light in The Great Gatsby” follow-up prompt to list sources. ChatGPT, 13 Feb. version, OpenAI, 9 Mar. 2023, chat.openai.com/chat.

While we’ve provided fairly detailed descriptions of the prompts above, a more general one (e.g., Symbolism of the green light in The Great Gatsby prompt) could be used, since you are describing something that mimics a conversation, which could have various prompts along the way.

Example 3: Citing Creative Visual Works

If you are incorporating an AI-generated image in your work, you will likely need to create a caption for it following the guidelines in section 1.7 of the MLA Handbook. Use a description of the prompt, followed by the AI tool, version, and date created:

A pointillist digital painting of a sheep in a sunny field of blue flowers

Fig. 1. “Pointillist painting of a sheep in a sunny field of blue flowers” prompt, DALL-E, version 2, OpenAI, 8 Mar. 2023, labs.openai.com/.

You can use this same information if you choose to create a works-cited-list entry instead of including the full citation in the caption (see MLA Handbook, sec. 1.7).

Example 4: Quoting Creative Textual Works

If you ask a generative AI tool to create a work, like a poem, how you cite it will depend on whether you assign a title to it. Let’s say, for example, you ask ChatGPT to write a villanelle titled “The Sunflower” that—you guessed it!—describes a sunflower and then quote it in your text. Your works-cited-list entry might look like this:

“The Sunflower” villanelle about a sunflower. ChatGPT, 13 Feb. version, OpenAI, 8 Mar. 2023, chat.openai.com/chat.

If you did not title the work, incorporate part of or all of the first line into the description of the work in the Title of Source element:

“Upon the shore . . .” Shakespearean sonnet about seeing the ocean. ChatGPT, 13 Feb. version, OpenAI, 8 Mar. 2023, chat.openai.com/chat.

For guidance on using descriptions and text from the work itself in the Title of Source element, see the MLA Handbook, 5.28 and 5.29.

Example 5: Citing Secondary Sources Used by an AI Tool

You should also take care to vet the secondary sources cited by a generative AI tool—with the caveat that AI tools do not always cite sources or, when they do, do not always indicate precisely what a given source has contributed. If you cite an AI summary that includes sources and do not go on to consult those sources yourself, we recommend that you acknowledge secondary sources in your work.

For example, let’s say that you ask Bing AI to explain the concept of the political unconscious, citing sources, and it provides the following answer:

Screenshot of Bing AI response about the political unconscious

Let’s say that you then decide to quote from the final sentence. You need to click through to the source listed in the note in order to get more information than just a URL for the source. There, you will read the following:

Screenshot of Oxford Reference web page about the political unconscious

Now, you can treat Oxford Reference as your source since Bing AI was merely a research conduit to the source (see MLA Handbook 5.34 for more information). If for some reason you want to treat a source cited in a generative AI tool as an indirect source–and you know it is, in fact, the source for the information provided by the AI, follow the guidance in section 6.77 of the MLA Handbook.

Note

1. At the time of writing this post, ChatGPT doesn’t have a built-in feature to create a unique URL to the conversation. However, an outside tool like the Chrome extension ShareGPT can generate such a link. If you use that type of outside tool, include the unique URL that the tool generates instead of the general URL.

DALL-E allows users to download the AI-generated images they create or generate a publicly-available URL that leads to an image. If you choose to create a shareable link for an image you generate with DALL-E (or other similar AI image generators), include that unique URL that leads to the image instead of the general URL.

Works Cited

MLA Handbook. 9th ed., Modern Language Association of America, 2021. MLA Handbook Plus, 2021, mlahandbookplus.org/.

Weed, Julie. “Can ChatGPT Plan Your Vacation?” The New York Times, 16 Mar. 2023, www.nytimes.com/2023/03/16/travel/chatgpt-artificial-intelligence-travel-vacation.html.