AI Tools for Academic Literature Research

Balancing Potential Gains and Losses

"There's no way I have time to keep up with all the literature published in my discipline!"

AI literature research tools can increase your efficiency by

• Helping you find articles/chapters, etc. on a topic and providing related works and authors 
• Offering a mechanism for quickly identifying seminal and frequently cited works
• Tracing citations forward and backward in a collection of related works; viewing a timeline of scholarship related to a topic
• Summarizing and/or extracting key elements or points from articles
• Potentially providing access to information in other languages that you would otherwise not attempt to use 

But there are also downsides.

AI literature research tools have some shortcomings, especially the "freemium" tools

• They do not have access to every database, collection, etc. For example, they may not be able to pull resources from behind paywalls.
• The algorithms may influence the resources recommended to you in unknown or negative ways. For example, they may be privileging certain scholars and works or groups of scholars and works over others.
• Paid applications tend to offer advanced features and can lead to better outcomes. For example, paid users of Elicit can add one "high accuracy" column. (Who wants low accuracy?!) Differences in output based on who can afford paid tools creates inequities. 
• Use of the AI summaries and information extractions poses risks of users'...
  • Overreliance on the AI tools and loss of interpretation and critical evaluation skills
  • Spread of false information if the AI output is not checked
  • Missing nuance and important contextual information in materials
  • Being predisposed to think a certain way about the information based on the AI-generated content

Generative AI Models Have Been Trained Using Content Collected from the Web

• Huge quantities of data have been required to train the generative AI models and systems that are available. A large percentage of that data was scraped or otherwise captured from the Web without asking creators' permission and without providing them with compensation (Kim, 2025). 
• In the United States, authors, musicians, media publishers, and others have filed lawsuits claiming copyright infringement against generative AI companies including OpenAI, Anthropic, Microsoft, Perpelxity, Meta, Suno, Udio, and others In most of these cases, the AI companies are arguing that their ingestion of the content falls under fair use and should be considered transformative use (Madigan, 2025). Some of the argument for fair use may rest on the issue of how generative AI (genAI) systems operate. While they ingest data in one form such as text, that data is typically converted into machine readable data and processed some more before it is then output again in text, audio, or video format. For more on genAI systems and how they work, see Laubheimer (2024) and Zewe (2023). 
• The U.S. Copyright Office has been preparing a 3-part report on Copyright and AI. Part 1 deals with protecting people from "unauthorized digital replicas" (e.g. deepfakes) and was published in July 2024. Part 2 addresses "copyrightability" of works that include AI-generated content. It was released on January 29, 2025--Some works that were created with AI assistance can be copyrighted depending on the level of human input, decision-making, assemblage, etc.. Part 3 will tackle the issue of data ingestion from online sources for use in training of genAI systems. See the U.S. Copyright Office website "Copyright and Artificial Intelligence." 

Helpful Links

Other Ethical Concerns Associated with Widespread AI Use

This guide is focused on research-related AI tools, so it includes limited detail about broader ethical issues. A few are noted below with sources to initiate further exploration:

Biased data and algorithms used in AI systems. These can have disproportionately negative impacts on already marginalized groups. 

• Stephenson, B., & Harvey, A. (2022). Student equity in the age of AI-enabled assessment. In R. Ajjawi, J. Tai, D. Boud, & T. Jorre De St Jorre, Assessment for Inclusion in Higher Education (1st ed., pp. 120–130). Routledge. https://doi.org/10.4324/9781003293101-14

• Noble. S. U. (2018). Algorithms of oppression: How search engines reinforce racism. New York University Press.

• O'Neil, C. (2016). Weapons of math destruction: How big data increases inequality and threatens democracy. Crown

Risks to privacy and potential for over-surveillance 

• Briggs, M., & Cross, M. (2024). Generative AI: Threatening established human rights instruments at scale. 2024 4th International Conference on Applied Artificial Intelligence (ICAPAI), 1–8. https://doi.org/10.1109/ICAPAI61893.2024.10541170

• Jones, K. M. L. (2022). The datafied student—Why students’ data privacy matters and the responsibility to protect it (pp. 1–18). Future of Privacy Forum. https://studentprivacycompass.org/wp-content/uploads/2022/04/FPF_Jones-Research-Brief_R3.pdf

Invisible laborers enabling AI --Considering data labelers & others behind the scenes

• Quinn, A.-E. (2024, December 11). Updating human rights law necessary to combat “digital forced labor” in age of GenAI. Thomson Reuters Institute. https://www.thomsonreuters.com/en-us/posts/human-rights-crimes/digital-forced-labor/

• Sarkar, A. (2023). Enough with “human-AI collaboration.” Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems, 1–8. https://doi.org/10.1145/3544549.3582735

Environmental impacts of data processing and computing required for AI 

• Zewe, A. (2025, January 17). Explained: Generative AI’s environmental impact. MIT News. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117

• Ren, S., & Wierman, A. (2024, July 15). The uneven distribution of AI’s environmental impacts. Harvard Business Review. https://hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts