How to Cite AI Tools

Our last two blog posts have been about citing/acknowledging AI tools. This is the third (and final, for now) installment; another contribution by Dr. Kristin Terrill, who is the Graduate Student Services Specialist at Iowa State University’s Center for Communication Excellence. For more context, please see our original post about acknowledging AI Tools, and her first response, “Should I Cite AI Tools?”


How should I cite the AI tool that I used?

If you used an AI tool like those developed at Moxie (formerly the Academic Insight Lab), Grammarly, or ChatGPT, cite it following your citation style’s (e.g., APA, MLA, etc.) guidelines for software. This will entail some fact-finding. You’ll need to know:

  • The author of the software (software authors are often corporations, not individual people)

  • The full title of the software

  • The version that you used

  • The date that version was published

Unlike books and articles, this information is not always helpfully presented front and center with software. If the software is running on your computer, you might find an “About” option in one of the application’s menus. If you downloaded the software or installed it from a flash drive or CD-ROM (yes, I’m that old), then there may be a separate “About” file in the program folder. If the software is a web application that you access through your browser, documentation may be provided in a menu available on the app. You probably need to hunt around—web app interfaces are generally not designed to help academic researchers find relevant citation information. As is true for all human knowledge, a librarian can help you if you get stuck.

Including the LLM details in the citation for an AI “application layer” or “wrapper”

Recently I disagreed with my esteemed colleague, a head developer at Moxie, about how to cite software that integrates other software. Moxie apps use large language models (LLMs) as part of a more complex process. My colleague suggested that the specific LLM used should be included in the long citation that goes in the References list. I disagree, and I’ll explain why.

Most modern software is made of building blocks—i.e., other software. You have probably heard the term algorithm, and you may also be familiar with terms like package, subroutine, and API. These terms are all related to the concept of combining multiple computer programs into a single, more complex program. This is what “application layer” developers do, and they add value by making a basic program suitable for a specific task.

Citations for software don’t include details about the building blocks for that software; they only include the version number and/or publication date. Readers can go find the documentation for that version themselves and said documentation may (or may not) include details about the building blocks that were used for that version. Just as some chefs won’t reveal their secret ingredients, a lot of developers stay mum about their algorithms.

Now, if you are citing the software, that means you used it as part of your research methodology, so you do need to provide relevant information about how it works—this information goes in the Methods section! If you get to choose which LLM you want to use, as some AI application layer software allows, then this information should be clearly presented as part of your methods. Otherwise, just point the reader to the information provided by the software publisher, and you will have given them the information they need.

 

Acknowledging the use of LLMs for writing assistance

Those who use chatbots to generate content for research articles may feel ambivalent about taking credit for the work. That’s fair, and remember, it is a controversial subject. Some authors deal with this ambivalence by including an ‘Acknowledgments’ section in their article. This kind of article section traditionally provides authors with space to tell the story of how they wrote the article and give credit to people who helped in a meaningful way but whose contribution didn’t rise to the level of authorship. This is a wonderful professional courtesy and a way of conveying your humility as an author. In my opinion, it is not inappropriate to include a nod to an LLM in the acknowledgments. However, it still doesn’t absolve you, the author, of your full responsibility for the content of the article. I think it’s debatable whether LLMs deserve the honor of being named in an Acknowledgment; I’m on the fence about this, myself. If you don’t want to “go public” as an LLM user but want to be transparent with the publisher, you can skip the Acknowledgments and simply explain in your cover letter (semi-private correspondence between you and your publisher) about how you used LLMs in your writing process. Similarly, if it’s a thesis/dissertation that you used an LLM to help write, you can convey this to your exam committee in a separate email or letter of transmittal.

Takeaways

As you integrate AI tools into your academic workflow, adhere to these key principles:

  • Software Citation Matters: Cite any AI tools used just as you would other software, following your selected citation style format. Be sure to note the corporate author, if applicable, along with the full title, version, and publication date.

  • Transparency in Systematic Reviews: Disclose the use of AI literature search and screening tools in systematic reviews and meta-analyses where methodology is integral to interpreting results. However, it is generally unnecessary to cite AI assistance with informal literature reviews.

  • Acknowledge AI Writing Support: If you received AI writing assistance, consider including an acknowledgment. However, this does not relinquish your responsibility as author to verify accuracy and facts. Citation implies verification; an acknowledgment does not.

  • Uphold Accountability: While AI tools can provide helpful support, you remain accountable for the final published content. Do not overstate their involvement or rely upon them to catch all errors. Maintain responsibility for fact-checking, logical flow, and scholarly integrity.

By following these recommendations, you can appropriately incorporate AI tools within a rigorous, ethical academic workflow.

For another take on this topic, please see chapter 13 (Acknowledging and Citing Generative AI in Academic Work) in the e-book, Write What Matters.

Elevate Your Research with AI

Dive into a curated collection of AI tools specifically designed to enhance and streamline your academic journey.