What Effective AI Legal Research Will Look Like

Like everyone else, I have watched several presentations this year on how AI is being incorporated into legal research platforms. Much of what I have seen is generative AI being used in a way that has greater implications for legal drafting than for legal research. In terms of AI-powered research, everything comes with the caveat of “Don’t trust, verify”—meaning it isn’t yet a game changer. The need for quality legal research instruction and assistance has not eliminated. I still plan to teach the same principles and techniques to students.

Even so, as I explore the AI tools that have been built so far, I have considered the contours of a sophisticated, capable AI legal research tool. What will a game-changing AI research tool look like? A truly useful legal research product will help me drill down more quickly to the resources I need but ultimately leave me to do the legal analysis. This is what I think an effective AI legal research tool will do:

1. Create accurate, custom legal encyclopedia or treatise articles. Not all states have rich collections of secondary sources or encyclopedias, and not all topics are covered in existing secondary sources. Game-changing AI will identify and summarize the most important primary sources on a topic much like a treatise or encyclopedia article.
2. Develop meta-analyses of secondary sources. For those of us lucky enough to have access to multiple secondary sources, the AI-powered research tool will gather and synthesize these sources into a single analysis with relevant citations and passages from the original sources, similar to a meta-analysis of scientific or medical studies. This functionality will be especially valuable when putting together a policy argument for a brief or a literature review for an academic article.
3. Produce very specific results for case law searches. Powerful AI will take natural language searching to the next level, returning relevant results to more complex searches via its ability to parse or summarize a document. For example, it can be difficult to search for cases that provide detailed analysis for a particular prong of a legal test, but AI-supported searching will be able to do just that.
4. Generate a timeline of important legal sources and events. I envision this as a sort of “super citator” that highlights changes in the law over time. For example, if I am researching a problem that involves a statute and associated case law, I could ask the system for a timeline showing the initial implementation of the statute, interpretative cases with subsequent treatment, and any amendments to the statute, with summaries for each. 
5. Allow for flexible searching. A useful AI research system will allow for both natural language searching and Boolean searching in combination and at different points in the search process. Although the system will not demand complex prompt engineering skills, it will require a well-developed research question.
6. Act as a creative sandbox. Due to the nature of computing technology, traditional research platforms are very literal. Electronic research does not provide the same serendipity that occurs when browsing titles in a physical library or skimming through a stack of books. An ideal research tool will include both literal research spaces and experimentation spaces.

An AI-powered system with these capabilities would be a game-changer, reducing the tedious sorting through cases and secondary sources and getting attorneys and librarians more quickly to the important work of legal analysis.