Guidance about whether or not and how to properly cite text generated by chatbots is still evolving. Authors often work with a range of technologies, and a chatbot is potentially a useful technology; however, since chatbots are not people, they are generally not recognized as authors. Unlike human-generated language, chatbot output is not based on reasoning, logical processes, or knowledge of the (social) world. Instead, chatbots generate probable word sequences, much as autocomplete suggestions for completing a text message. The import of the chatbot’s text depends on the human taking responsibility by asserting authorship over and through these words. 

Chatbot-generated text is often not reproducible, even with the same prompt. In academic contexts, it is therefore important to document what was generated so that readers have access to it. A general best practice is that chatbot content should be noted in the paper along with details about how it was used and the prompt (request) and response - see Committee on Publication Ethics. 

In many cases, journals and citation style guides are just starting to detail their guidelines for handling GenAI content. Here are guidelines from a few of them:

  • APA Guidelines:

    • Recognizing/Crediting Text: Since responses generated by chatbots are not viewable by others, APA considers the content to be treated as non-retrievable data or quotations. APA does not currently recommend citing it as “personal communications” since no person is involved. 

      • In-text Citation: APA’s current guidance is to credit the algorithm with an in-text citation (e.g., Open AI, 2023) and an entry in the reference list. 

      • Reference Example: APA suggests sometime like this: OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat. {Note that ownership of text output is an open question.}

      • Acknowledging Use of Chatbots: When chatbots or GenAI are used for research, authors should describe how they used the technology in their methods section. If this/these technologies are used in the literature review, then authors should describe in the introduction how the tool was used. The prompt should also be provided along with any of the relevant text generated by the chatbot.

  • MLA Guidelines

    • MLA recommends that GenAI tools should be cited whenever their content is “paraphrase(d), quote(d), or incorporate(d)” into an author’s work. MLA also recommends that any “functional uses” of GenAI, including editing or translation technologies, be acknowledged in a note or in the text, or in another appropriate location in the text. Since chatbots can produce fabricated or falsified sources, MLA also asks authors to check any secondary sources used or cited by chatbots. 

      • In-text Citations:

        • Chatbot content should be referenced by the prompt that generated it. This prompt is included in the in-text citation that identifies paraphrased or quoted material. See MLA Guidelines for examples. 

      • Works-Cited-List Entry:

        • Based on the MLA Template model, any chatbot content should be included at the end of a piece and should have the Title of the Source, the Title of the Container, the Version, the Publisher, the Date, and the Location. Note that since MLA does not view chatbots as authors, the first piece of information in the works-cited-list entry will be the Title of the Source, in other words, the prompt used to elicit the content. For example, 

          • “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.

    • Since MLA does ask for a URLfor the source, writers should consider using a program that allows for generating URLs for prompts and associated responses. There are at least two programs that generate this information - A.I. Archives and ShareGPT (a Chrome extension).

  • The Chicago Manual of Style (CMOS)

    • CMOS recommends the text or content generated by a chatbot needs be acknowledged, even if the text was edited by the author and that chatbots must be credited when its text is “reproduced” within a writer’s text, but that the information should be included either in the text or in a note but not included in a bibliography or reference list. CMOS recommends that any prompts used to elicit responses from a chatbot be included in the text or in the footnote/endnote. 

    • Writers should use a similar approach when using images, figures, etc., generated by GenAI

      • In-text Citations: 

        • For less formal uses, authors can note in text something like “the following content was generated by Bard”;

        • For more formal citations (such as a footnote or endnote), authors should include that the text or material was generated by a chatbot (i.e., the author), along with the date and publisher or sponsor of the content - for example, “Text generated by ChatGPT, August 8, 2023, Open AI, https://chat.openai.com/chat.

      • Bibliography / References Section: CMOS currently (as of August 2023) does not consider it appropriate or necessary to include chatbot citations in this section of a writing piece.