Skip to Main Content

Generative AI and ChatGPT

When and What to Cite from Generative AI

While some citation and referencing styles have published interim direction on how AI generated materials should be cited, many still do not have specific guidelines. In general, though, you should treat AI-generated materials as a non-recoverable source and/or akin to personal communication, and follow guidelines for how to cite those resource types with your chosen reference style.

Below are some suggestions on citing generative AI materials in some of the major citation styles. Please note that UBC Library has a more comprehensive guide on how (and when) to cite other resources here. As always, please be sure to check with your instructor if in doubt about when to cite AI and what style to use. 

Citing AI Generated Materials

In a post from April 7, 2023, APA offers some guidance on citing AI-generated materials and AI software.

First, APA suggests ensuring you describe how you used the AI tool in your research in a method section or comparable section of your paper.

APA also suggests that given that AI generated content like chats are not created by a person, that they cannot be considered personal communication. Instead, treat the content as an algorithm's output, and credit the author of the algorithm with a reference list entry and in-text citation.

You can also put the full text of long responses in an appendix or online supplemental materials.

 
Reference List
Author. (Year of the Version). Title of the Tool (Version if applicable) [Tool description if applicable]. Source/URL

Example:

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

 
In-Text Citation

For a direct quote or paraphrase, or to cite the tool, use the following format. Note: OpenAI is the author not ChatGPT.

(Author, Year)

Example:

Parenthetical citation: (OpenAI, 2023)

Narrative citation: OpenAI (2023)

As of March 17, 2023, MLA has provided some guidance on citing generative AI, including ChatGPT and DALL-E.

In summary, MLA recommends

  • citing a generative AI tool whenever you incorporate any content created by it into your own work
  • acknowledging any uses of the tool in a note or other suitable location
  • vetting any secondary sources it cites
  • not treating the AI tool as an author

 

Works Cited Entry

The works cited entry for AI generated material should follow this format:

"Description of chat" prompt. Name of AI tool, version of AI tool, Company, Date of chat, URL.

Example:

"Write a haiku in the style of Edgar Allen Poe" prompt. ChatGPT, May 3 version, OpenAI, 19 June 2023, chat.openai.com/chat

 

In-Text Citation

For any quotes, paraphrases or references to the AI generated material, use the following format:

("Shortened description of chat")

Example:

("Write a haiku")

The Chicago Manual of Style does not have official recommendations for reference generative AI, but does provide some guidance about citing AI generated materials in a Q&A. (as of June 2023).

Currently, Chicago treats the AI model as an author, and the format as a personal communication; therefore references to AI generated content are included in a footnote but not in a bibliography or reference list.

 
Bibliography or Reference List

Do not include.

 
Footnote Version
Note number. Originator of the communication and description of the prompt, date the text was generated, publisher, General URL

Example:

1. ChatGPT, response to "Create a haiku in the style of Edgar Allen Poe," June 16, 2023, OpenAI, https://chat.openai.com/chat.

 

Author-Date Version

In an author-date version where parenthetical citations are used in text, any information not in the text would be placed in a parenthentical reference.

Example:

(ChatGPT, June 16, 2023).

Although there is no specific guidance offered by Vancouver Style editors, AI content is generally considered personal communication and/or non-recoverable data, and therefore should not be included as a reference in your reference list according to the Vancouver Style.

 
Reference List

Do not include.

 
In-Text Citation

AI-generated content can be cited in-text using the following format:

Type of Communication, Communicator, Date (DMY)

You can also include further information, such as the prompt you offered the AI in type of communication.

Examples:

In an online chat with Open AI's Chat GPT (23 March 2023)...

A chat generated a potentially useful script for dealing with these interactions (ChatGPT response, prompt for "How to resolve conflict with a coworker", 24 March 2023).

...(Grammerly paraphrase, 22 February 2023).

As of June 2023, IEEE does not cite AI generated text for publication and does not consider AI generated text a valid reference.

However, overall current advice is to treat AI generated text as a private communication or nonrecoverable material. Therefore, do not include a citation number, or include the reference in your reference list.

 

Reference List

Do not include a reference list entry.

 

In-text Citation

IEEE uses the following in-text citation format for unpublished materials:

Author's name (Initials, Surname), private communication, Abbrev. Month, year.)

Example:

In an online chat,... (OpenAI's Chat GPT, private communication, June 2023)

As with textual outputs, it is important to acknowledge how you have used generative AI in creating non-textual outputs such as images, music, figures, etc.

As of June 2023, only MLA and Chicago have released some guidelines on how to cite visual and other works generated by AI. See the links below for more information on citing images and other non-textual materials generated by AI, as well as guides created by UBC Library on citing non-textual materials generally.

AIs and Authorship

As of June 2023, several publishers have offered guidance on how AI tools should be considered for publications

  • Cambridge University Press states that AI must be declared and clearly explained, and does not meet the requirements for authorship in terms of accountability (Link to Cambridge Authorship and Contributorship Guidelines)
  • Taylor & Francis states that AI tools cannot be listed as an author, and their use must be acknowledge and documented appropriately (Link to Taylor and Francis statement on responsible use of AI)
  • Nature and other Springer Journals advise that Large Language Models (LLMs) like ChatGPT will not be accepted as credited authors on research papers. However, researchers using LLM tools should document their use in the methods or acknowledgement section (Link to Nature statement on use of AI)
  • Science has created an Artificial Intelligence (AI) policy which states: “Text generated from AI, machine learning, or similar algorithmic tools cannot be used in papers published in Science journals, nor can the accompanying figures, images, or graphics be the products of such tools, without explicit permission from the editors. In addition, an AI program cannot be an author of a Science journal paper. A violation of this policy constitutes scientific misconduct.” (Link to Science statement)

Technical Standards

The Coalition for Content Provenance and Authenticity is developing an open technical standard for greater transparency in how media or documents have been altered. The MIT Technology Review has described this as a "nutritional label" for AI.