No CrossRef data available.
Article contents
AI-Assisted Legal Research: AIALR
Published online by Cambridge University Press: 29 August 2025
Abstract
The rise of GenAI (generative AI) tools such as ChatGPT has transformed the research environment, yet most legal researchers remain untrained in the theory, mechanics and epistemic structure of such systems. The public itself was introduced to GenAI through Generative Pre-Trained Transformer tools such as ChatGPT and Claude. Although AI is a decades-old academic discipline, it is now rapidly expanding, and LLM-based AI tools (called Legal Research AI Tools, or LRATs hereinafter, such as Lexis+ AI) sit at AI’s cutting edge within legal research. These LRATs rely on non-legal theoretical informational concepts and technologies to function. Legal researchers often struggle to understand how AI-enabled tools function, which makes effective/reliable use of them more difficult. Without proper orientation, legal professionals risk using LRATs with misplaced confidence and insufficient clarity, the implications of which will be addressed in a future article. This article, written by Ryan Marcotte, Reference, Instruction, & Scholarship law librarian at DePaul University’s College of Law in Chicago, Illinois, defines and explains AI-assisted legal research (AIALR) as a third phase of research logic following the traditional book-based legal research (BLR) and computer-assisted legal research (CALR) phases. It also introduces a definition of AI tailored for legal research, outlines key conceptual structures underpinning LRATs, and explains how they interpret human input. From this grounding, this article offers two frameworks: (1) the Five Ps Research Plan and (2) the four prompt engineering methodologies of Retrieval Augmented Generation, Few-Shot Prompting, Chain-of-Thought/Chain-of-Logic, and Prompt Chaining. Together, these frameworks equip legal researchers with the understanding and skills to plan, shape, and evaluate their research interactions with LRATs in the age of GenAI.
Keywords
Information
- Type
- Main Features
- Information
- Copyright
- © The Author(s), 2025. Published by British and Irish Association of Law Librarians
References
Endnotes
1 I would like to thank the staff at DePaul University College of Law’s Rinn Law Library, particularly Marcos Corona, for their aid in ensuring that the conceptual analogies in this article make sense to legally trained minds. I would also like to thank Jill Marcotte, Rachel Riffe, and Kevin Wiggins for their assistance in ensuring that they made sense to non-legally trained minds.
2 Jamie J. Baker, A Legal Research Odyssey: Artificial Intelligence as Disruptor, 110 Law Lib. J. 7 (2018)
3 Staff of H. Comm. on Fin. Servs., Potential Impacts of Artificial Intelligence in Financial Services, 118th Cong. (Sept. 2023), https://democrats-financialservices.house.gov/uploadedfiles/ai_report.pdf.
4 LexisNexis, AI Goes to Law Schools, Continuing Legal Education Program (2023).
5 C.M. Poulson, FlightGlobal, Radar for Airlines, p. 434 (02 May 1945), <https://web.archive.org/web/20190214002854/https:/www.flightglobal.com/pdfarchive/view/1946/1946%20-%200844.html> accessed 7 February 2024.
6 Stanley Greenstein, ‘Preserving the Rule of Law in the Era of Artificial Intelligence’ (2022), 30 Artificial Intelligence and L 291, 299.
7 IBM, ‘AI vs. Machine Learning’ (YouTube, Apr. 10, 2023) <https://youtu.be/4RixMPF4xis?si=fNVFuWP2MdesIRey> accessed 18 July 2024.
8 Terrence J. Sejnowski, The Unreasonable Effectiveness of Deep Learning in Artificial Intelligence, 117 PNAS 48 (2020).
9 Rebecca Slayton, ‘The Promise and Risks of Artificial Intelligence: A Brief History’ (War on the Rocks, 08 June 2020) < https://warontherocks.com/2020/06/the-promise-and-risks-of-artificial-intelligence-a-brief-history/> accessed 12 July 2024.
10 e.g., TurboTax tax software, H&R Block tax software.
11 Sejnowski (n 8).
12 ibid
13 ibid
14 Daniel E. O’Leary & Robert M. O’Keefe, ‘The Impact of Artificial Intelligence in Accounting Work: Expert Systems Use in Auditing and Tax’ (1997) 11 AI & Society 36.
15 Enrico Francesconi, ‘The Winter, The Summer, and The Summer Dream of Artificial Intelligence in Law’ (2022) 30 Artificial Intelligence & Law 147.
16 J.R. Searle, Is the Brain’s Mind a Computer Program, 262 Sci. American 20 (1990).
17 Marvin Minsky, The Society of Mind (Simon & Schuster, 1986).
18 Greenstein (n 7).
19 IBM, ‘Neural Networks Explained in Five Minutes’ (YouTube, 24 May 2022), <https://youtu.be/jmmW0F0biz0?si=1JYPAFCzNZllnUJh> accessed 4 September 2024.
20 Greenstein (n 6).
21 ibid
22 ibid
23 IBM (n 19).
24 Greenstein (n 6).
25 William Shakespeare, Romeo and Juliet Act 2, sc. 2, l. 46-7.
26 ibid
27 Greenstein (n 6).
28 ibid
29 Susan Nevelow Mart, Every Algorithm Has a POV, AALL Spectrum, Sept.-Oct. 2017 at 40.
30 ibid
31 Baker (n 2).
32 Greenstein (n 6).
33 IBM, What Is NLP? (YouTube, 11 August 2021), < https://youtu.be/fLvJ8VdHLA0?si=ua5aD-7DCYH9xXdt> accessed 20 March 2025.
34 Craig W. Schmidt, et al., Tokenization is More Than Compression (2024), <https://doi.org/10.48550/arXiv.2402.18376> accessed 22 March 2025.
35 IBM (n 19)
36 Itô, Kiyosi (1993). Encyclopedic Dictionary of Mathematics (2nd ed.).
37 
38 IBM (n 19).
39 Louie Giray, Prompt Engineering with ChatGPT: A Guide for Academic Writers, 51 Annals of Biomed. Engineer. 2629 (2023).
40 ibid
41 ibid
42 There are certain Boolean search syntax differences between Westlaw Precision and Lexis+; for instance, Westlaw prefers proximity searching (searching for Term X within a certain distance from Term Y) using “/X,” whereas Lexis prefers proximity searching using “W/X”.
43 Susan Boland, ‘Sources of Law and Legal Authority Video’ (University of Cincinnati Mediaspace, 20 June 2017) < https://uc.mediaspace.kaltura.com/media/Sources+of+Law+and+Legal+Authority+Video+--+by+Susan+Boland/1_vm6lb4r3> accessed 07 July 2024.
44 ibid
45 ibid
46 ibid
47 ibid
48 When one considers the analogy of filling in the section on a Venn Diagram where both X and Y exist, the section filled will include only the area of intersection where X and Y cross over into each other. Likewise, the section on a Venn Diagram where either X or Y exists will include the X-only section, the Y-only section, and the area of intersection of X and Y. Therefore, when a legal researcher considers Venn Diagrams in the context of Boolean logic gates, that researcher will recognize that “AND” is narrowing language as well as that “OR” is broadening language. “NOT” is by its inherent nature narrowing language.
49 These axioms are (1) that whichever terms are put inside of parentheses [represented by the X in the Boolean search string (X) & Y] are searched for before whichever terms are put outside of parentheses, (2) that whichever terms put within quotation marks are searched for exactly as they are written [represented by the X in the search string “X” & Y (3) that whichever search terms are located on the left side of a search string are searched for before whichever search terms are located on the right side of a search string due to computers reading searches from left-to-right [i.e., in the search string X & Y, X will be searched for before Y], and (4) that the combination of several search terms into a comprehensive search string can be achieved through the nesting of parenthetical search phrases combined with the usage proximity and/or truncator language.
50 e.g., The One Good Case Method [find one good case on a topic, and look back, forward, and around for other good cases using citators and headnotes].
51 “Epistemology, N.” Oxford English Dictionary, Oxford UP, July 2023, <https://doi.org/10.1093/OED/3496453927> accessed 20 April 2025.
52 Using the “Five W’s” mental framework of Who, What, When, Where, and Why to analyze one’s legal research issue such that they can plan their research, which one then translates into Boolean search language to search online legal research databases for legal materials that would help to answer the question at issue.
53 On the hypothetical question “Does the Stored Communications Act (SCA), part of the Electronic Communications Privacy Act (ECPA), 18 U.S.C. §§ 2701-2712., apply to emails held overseas by a U.S. cloud provider,” during the Prime P, the researcher would infer the legal focus is on the SCA, extract the prime facts that (1) the email data is stored overseas, (2) the data is stored by a U.S.-based cloud service provider, and (3) the issue here is extraterritorial application of U.S. law
54 On the same hypothetical question, during the Persona P, the researcher could infer that the question is likely a government-side question posed to assess compliance under SCA, which would affect which sources might be relevant to questions of that side, such as Dept. of Justice guidelines, seminal cases such as U.S. v. Microsoft Corp., 584 U.S. 236 (2018), and Fourth Amendment concerns, resulting in the researcher mentally weighing their cases and resources to prioritize government-side interpretive stances. This is an example of a stance a researcher could take during the Persona P, and the logic of this P could flip to search for non-government side resources.
55 Again, on the same hypothetical during the Prompt P, the researcher could infer that the question itself contains latent ambiguity, in that “data held overseas” could trigger irrelevant international privacy law when searched for unless the jurisdictional scope of the question is clarified by proper framing. After considering the Prompt P, the researcher could rearticulate their research question as “Under the SCA, may a U.S. law enforcement agency compel a U.S.-based cloud service provider to produce emails stored on servers located abroad” so that less search ambiguity remains.
56 During the Polish P, the researcher could acknowledge that the Clarifying Lawful Overseas Use of Data (CLOUD) Act was passed in direct response to Microsoft, and revise their research into the precise legislative text showing § 2713’s extraterritorial reach or OLC guidance on the U.S. government’s obligations to notify other, non-U.S. governments
57 Again, on the same extended hypothetical, during the Product P, an LRAT would (1) retrieve seminal resources, such as U.S. v. Microsoft Corp or In re Warrant to Search a Certain E-Mail Account Controlled & Maintained by Microsoft Corp., 829 F.3d 197 (2d Cir. 2016) or 18 U.S.C. § 2713, or secondary sources that may be called for by the direction of the research as informed by the Persona P, such as DOJ white papers or major law review articles, and (2) generate an output tailored for a federal legal memo, in this example.
58 “Most Triumphant Pledge of Allegiance (Bill & Ted Style): I pledge my totally excellent vibes to the righteous flag of the United States of Whoa-merica, and to the gnarly Republic for which it totally stands— one awesome nation, under the Big Dude in the sky, indivisible, with liberty and rockin’ justice for all the most excellent dudes and babes,” generated by ChatGPT 19 August 2024.
59 “In X State, should No. 1, No. 2, or No. 3 happen? Draft your response using only the information available in X state that I give you.” Another example of this could read “Given X contract, should I include A, B, or C clause? Draft your responses using this particular contract only.”
60 “Is a party restricted from contesting the validity of the counterparty’s ownership of intellectual property or otherwise bringing a claim against the counterparty for matters unrelated to the contract? I will give you examples of how to run this search, which are the following: [these would be the few shots to teach the LRAT]”.