Citing generative AI in APA Style: Part 3—Is AI “allowed” in APA Style?

Since the APA Style team wrote our 2023 post about citing ChatGPT, the landscape of generative artificial intelligence (AI) has changed dramatically. You have likely seen, and probably used, other AI toolsopens in new window in the meantime, and using these tools for assistance in the writing process may no longer seem unusual.

Although it’s still not a good idea to submit AI-generated work as though you did it yourself—doing so would be considered unethical, for example, academic misconduct for students—there are many potentially legitimate ways to use AI tools within the context of academic writing. For example, some authors—be they students, researchers, reviewers, instructors, or others—are using AI tools to search for information, summarize literature in conjunction with reading it, edit or analyze their own writing, brainstorm ideas, translate text, proofread for consistency and style, generate tables and figures, create and refine code, and much more. However, the details of what AI use is allowed and how it should be disclosed depend on many factors, such as instructor, departmental, and publisher policies. And of course, it is possible to cite AI tools without using them, such as if you are writing a paper about AI.

Because there are many issues to explore surrounding AI, we have split our advice into different posts. This will also allow us to add additional posts as needed in the future. This post is about whether AI use is “allowed” in APA Style.

Additional posts are available on these topics:

• How to cite AI chats and AI tools generally in APA Style
• AI as a search engine and AI integrated into common software

Opinions vary regarding whether students, researchers, and others should use AI to help with writing. Sometimes it’s a personal choice, but when writing for class, for a thesis or dissertation, or for publication, authors need to consider the policies and guidelines of their instructor, department, editor, publisher, or other policy-setting authority. That is, whether you can or should use generative AI in your writing depends on who you are writing for and what their AI-related guidelines or policies are.

Furthermore, the answer of whether AI is allowed is rarely “yes” or “no” but instead is “yes, for some AI tools, for some purposes, and no, for others.” Some instructors or policies may approve of using AI to generate ideas, summarize literature, check writing for missing information, correct grammar mistakes, make writing more concise, and more. Others allow use on the condition that students to justify it within the text of the paper, with the goal of helping students explore the ethical questions surrounding AI use. Others may prohibit any use or prefer authors to use specific AI tools that have specific qualities. Still others have come up with creative ways to lean into using AI to write, such as by asking AI to write a paper and then having students critique it. If your institution or publisher doesn’t have a set AI policy, you may also encounter differing opinions and recommendations and need to collaborate and communicate to determine the best path.

Many book and journal publishers do have guidelines for authors using AI. For example, if you are submitting a manuscript for publication in APA Journals, the APA policy on the use of generative AI in scholarly materials applies to you.

APA Journals’ policy regarding AI stipulates that

• AI cannot be an author (because, as noted in our post on citing AI in APA Style, AI is not a conscious human who can consent to the duties and responsibilities of authorship, which include responsibility for post publication changes such as corrections or retractions)
• the use of AI should be disclosed in the Method section, author note, or other relevant place in the text
• the AI tool(s) should be cited, with references in the reference list

The common theme is that you need to communicate with the people evaluating your work to determine acceptable uses of AI and the appropriate disclosure for readers and then provide citations and references that will be helpful.

You must carefully review all AI-generated text, verify its accuracy, and take responsibility for it before submitting or publishing it. Failure to undertake this due diligence undermines scientific integrity and the accurate communication of information. For example, an AI-generated summer reading listopens in new window recently made headlines when readers discovered that only five out of 15 titles on the list were real (Blair, 2025). Other notable examples of incorrect or imaginary AI-generated citations have cropped up in far more serious places, such as a recent “Make America Healthy Again” report from the U.S. Department of Health and Human Services and a book on machine learning published by Springer Nature (Aksenfeld, 2025).

The APA Style team has seen similar issues with source confabulation crop up when asking AI for references to support example statements: AI will confidently provide citations, but the citations are to works that are not real, or the works are real but do not actually support the statement. It is essential for human authors to retain responsibility for their work even when they use technology such as AI to assist in writing tasks. The effective communication of accurate and helpful information depend on it.

AI is also known for being sycophantic—in some cases relentlessly affirming, just as confident when it is right as when it is wrong (OpenAI, 2025). Even after correcting inaccuracies or biases in AI, it may continue being incorrect. Even just asking a question, such as how two concepts are related, may lead the AI to create an argument to support their relationship, when in reality, the two concepts have a far more tenuous relationship. If you had asked the opposite question or a different version of the question, you might have received a far different argument.

Because AI is emerging technology, research on its effects on writing (and thus thinking) has only just begun. Preliminary research from MITopens in new window on a project called “Your Brain on ChatGPT”opens in new window suggests that using AI to help with writing tasks, as opposed to using a traditional search engine or a “brain only” approach, is associated with decreased neural connectivity patterns and lower recall and sense of ownership of the information produced (Kosmyna et al., 2025). This makes sense to us—that people who have not thought deeply about the work they produced would be unable to engage with it deeply after the fact.

Although the ChatGPT-produced essays in the study had near-perfect structure and use of language, human raters in the study (who were unaware of any AI use and of the purpose of the study) also described the AI-produced essays as “soulless,” with no individuality or creativity (Kosmyna et al., 2025, p. 62). Bear in mind that these findings are preliminary: They do not warrant sensationalist conclusions like “AI is making people stupid.” Rather, they underscore that using human critical-thinking skills in the writing process, including interactions with AI, is more essential than ever. Those interested in developing their critical-thinking skills and avoiding sensationalism should also check out APA Style’s new Evaluating Online Sources Activity (PDF, KB)opens in new window.

In addition to ethical questions related to bias or misinformation, AI presents ethical questions related to data privacy, confidentiality, copyright, and intellectual property. Authors must carefully consider whether it is appropriate to put sensitive information into an AI, such as reviewers uploading a submitted manuscript to AI to analyze it for peer review. As before, AI cannot take the place of human peer review because AI is not human. Reviewers should check publisher policies to determine what AI use is appropriate in the review process.  

AI also presents ethical questions regarding the environment. AI requires immense computational power, which consumes considerable resources—possibly up to 10 times as much as a traditional web search (de Vries, 2023). Vast quantities of fresh water are needed for data-center cooling as well as thermoelectric or hydroelectric power generation (Pinheiro Privette, 2024). Other data centers that rely on fossil fuels, such as the xAI data centers in Memphis, Tennessee, pollute the air for miles, posing significant threat to human lives (Southern Environmental Law Center, 2025). It is wise to consider the impact of the use of AI on the broader world we all share.

Thus, as authors learn how to work with this emerging technology, it is essential that they apply the same level of scientific scrutiny to the output and interaction with AI and its positionality within the world context as they do to any other scientific or writing endeavor. AI was created by humans and largely trained on data generated by humans. For all of the great things humans have achieved, humans are also fallible and prone to bias; therefore, AI also has the potential to be both great as well as fallible and prone to bias. And even when fine-tuned for greatness, AI comes with real environmental costs. So, as with the question of whether AI is “allowed,” it is more a matter of abiding by the policies for your specific case and how people can best use it to improve human lives.

If you do use an AI tool, be sure to read the previous posts in this series for guidelines on whether and how to disclose and cite this use. Please let us know what AI-related questions you have in the comments. We also look forward to you sharing your thoughts and experiences with AI with the community so that we on the APA Style team can continue to make useable, helpful guidance on this topic.

Aksenfeld, R. (2025, June 30). Springer Nature book on machine learning is full of made-up citations. Retraction Watch. https://retractionwatch.com/2025/06/30/springer-nature-book-on-machine-learning-is-full-of-made-up-citations/opens in new window

Blair, E. (2025, May 20).  How an AI-generated summer reading list got published in major newspapers. NPR. https://www.npr.org/2025/05/20/nx-s1-5405022/fake-summer-reading-list-aiopens in new window

de Vries, A. (2023). The growing energy footprint of artificial intelligence. Joule, 7(10), 2191–2194. https://doi.org/10.1016/j.joule.2023.09.004opens in new window

Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. arXiv. https://doi.org/10.48550/arXiv.2506.08872opens in new window

OpenAI. (2025, April 29). Sycophancy in GPT-4o: what happened and what we’re doing about it. https://openai.com/index/sycophancy-in-gpt-4o/opens in new window

Pinheiro Privette, A. (2024, October 11). AI’s challenging waters. Center for Secure Water, University of Illinois Urbana-Champaign. https://cee.illinois.edu/news/AIs-Challenging-Watersopens in new window

Southern Environmental Law Center. (2025, April 9). New images reveal Elon Musk’s xAI datacenter has nearly doubled its number of polluting, unpermitted gas turbines [Press release]. https://www.selc.org/press-release/new-images-reveal-elon-musks-xai-datacenter-has-nearly-doubled-its-number-of-polluting-unpermitted-gas-turbines/opens in new window