White paper
The past, present, and future of legal research with generative AI
A steady stream of legal research innovations has helped lawyers find better answers more efficiently for over a century.
Since the 1800s, many innovations improved the accessibility of law and standardized the publication of court decisions. Later innovations, including indexing and classification systems, made searching for precedents in published cases faster and more comprehensive. Digitalization freed researchers from the confines of a physical library. Citation analysis systems allowed researchers to determine the current legal validity of a given precedent. Litigation analytics brought data-driven insights to legal strategy.
These innovations have built on each other as legal research has become more accurate and efficient.
Today, the introduction of generative AI (GenAI) to legal research is a giant leap forward for lawyers — but like earlier innovations, it does not stand alone. Generative AI dramatically improves a researcher’s “time to answer” by building on and integrating those previous innovations. GenAI will profoundly impact legal research, not just because of the strength of the underlying technology but because of the quality of data and the human-based enhancements the technology operates on.
“The problems legal researchers are solving haven’t changed significantly in the past thirty years,” says attorney and Thomson Reuters Law School Account Manager Mark Frongillo. “What has changed are the tools they have to solve the problems. When I practiced law and then began training future lawyers, we started with book research. Now, law professors, students, librarians, and associates are conducting research the way they think: by asking questions, considering fact patterns, filtering by jurisdiction — and trusting their research tool to help them identify and interpret documents that are on point.”
Below are eight stages of how legal research has become modernized throughout the years.
Stage one: Accessing the law
Innovation — access and standardization
The first step in the modernization of modern legal research was simply making source materials — particularly court decisions and statutes — available to a broader audience. Since the Middle Ages, various court reporters have recorded and published court decisions. As American jurisprudence matured in the 19th century, legal publishers like John B. West saw lawyers needing help to keep track of a higher volume of legal decisions and apply precedent in their jurisdiction. Since 1879, West’s National Reporter System has compiled cases from state and federal courts and organized them into reporter sets by jurisdiction.
The reporter sets grew, and law firms, corporate legal departments, law schools, judges, government agencies, and public law libraries amassed collections. They updated them as supplemental inserts arrived and devoted significant space to their care.
Standardization and widespread access were the innovations here. Soon, nearly every lawyer had access to a comprehensive set of authoritative court reports in their jurisdiction and beyond. Publishers like West added a layer of human editorial treatment that enhanced the quality of court documents and standardized their structure and organization by jurisdiction.
Stage two: Finding the law
Innovation — classification, analysis, and summarization
Simply making the law more available in a standardized format was not enough. The proliferation of court decisions created a new challenge: keeping up with the law in a given jurisdiction or finding just the right cases to address the precise legal issue the lawyer was researching.
On top of the body of published law, publishers built a layer of editorial enhancements that made it easier to find and interpret the law. For example, West developed the Key Number System as a master classification system for U.S. law. Attorney-editors read every published opinion and assigned each issue in the case to a Topic and Key Number, taking context from the decision. Using printed digests, lawyers could quickly identify cases that addressed a specific issue, narrowing their search by jurisdiction as needed.
Headnotes — another editorial enhancement — helped researchers understand and organize court decisions. Attorney-editors identified the principal legal issues in a case and then wrote a brief description summarizing the facts, holding, and reasoning applied so researchers could determine the case’s relevance. The headnotes helped lawyers understand a case's substantive and procedural issues to assess its relevance and value as precedence.
Other analytic tools, such as American Law Reports, organized commentary on legal issues in a way that helped lawyers understand the prevailing law and find the underlying case precedents. Volumes of statutes and regulations, law review articles, and topical treatises or practice guides also served this dual purpose of helping to understand the state of the law and to identify essential precedents.
The innovations developed in this era of legal research moved legal publishing beyond collection and publication and added a level of human intelligence to primary legal sources.
“Our editorial enhancements help attorneys to research more efficiently and to have a lot of confidence they understand the state of the law,” says Jim Scott, Director of the Current Awareness team at Thomson Reuters. “All the attorney-editors’ work is organized around supporting those two principles.”
Stage three: Digitalization
Innovation — 24/7 availability and new search capabilities
Digitization of legal research began in the 1970s with the introduction of online research systems like Westlaw. The legal content formerly locked up in books was now available for remote access from anywhere, any time. But this innovation did more than simply make book content more accessible; it added a new arrow to the researcher’s quiver. Where the West Key Number System classification scheme enhanced print research, researchers could now augment those techniques with searches for specific words, phrases, concepts, proper names, etc.
While early versions of Westlaw allowed researchers to work from more locations, save their research histories, and expand their search techniques, it still mimicked the book research experience. Researchers still had to know how the case law was organized to find the right area of law. The documents returned by the computer still had to be read and sorted to determine which ones were on point and good law. Research processes had to reflect the organization and structure of the underlying print sources, which wasn’t always intuitive or easy to learn.
However, the innovations were building on each other as the new online research tools leveraged all the editorial enhancements and finding tools that started in the print era.
Thus, Westlaw built even its earliest versions on the foundation of the editorial capabilities first appearing in the print era and augmented them with new research approaches.
Stage four: Natural-language search
Innovation — natural-language searching
The marriage of artificial intelligence and the human element took another giant step forward in 2010 when Westlaw introduced natural-language searching alongside the traditional terms-and-connectors approach. With natural-language searching across all of Westlaw, researchers could use their own words and Westlaw would deliver relevant results for their jurisdiction. The attorney-editors’ thorough approach to assigning Topics and Key Numbers and adding synonyms to headnotes enabled Westlaw to deliver more targeted and comprehensive search results — with less input and less effort from legal researchers.
These advances focused primarily on helping researchers find results related to their specific questions. What does the law say on this particular topic with this set of facts? Can I still rely on this case for this point of law? What have other attorneys written about similar cases?
The pivot to natural-language searching signaled the growing ability of machines to process large amounts of legal data and see patterns and connections within it.
Stage five: Automating with AI
Innovation — automation in the research process
In 2018, Thomson Reuters introduced new ways of working with the release of Westlaw Edge. “Westlaw Edge is full of breathtaking science, and the AI and machine-learning components build perfectly upon our previous innovations and exclusive legal expertise,” said Mike Dahn, Senior Vice President and Head of Westlaw Product Management. “With Westlaw Edge, legal professionals can conduct legal research more quickly than ever before and uncover valuable new insights, giving them greater confidence when filing court documents or advising clients.”
Critical artificial intelligence innovations included:
AI in citation analysis
An essential task for lawyers is determining the validity of a cited source. Traditional citator services only flagged direct citation relationships between cases. Those citators did not warn researchers when a point of law had been overruled by implication, which could leave them at risk of citing invalidated law. Westlaw Edge introduced KeyCite Overruling Risk, which uses AI to identify cases that implicitly overrule or invalidate another case. This feature can be necessary, for example, when a court cites a case as authority, and that cited case is then overruled in a subsequent decision without explicitly overruling the original cited case.
AI in document analysis
Technological advances have also made it possible to leverage AI when analyzing briefs. Quick Check allows researchers to upload a brief and receive an in-depth report with recommendations for additional relevant authority, warnings for cited authority, an analysis of quotations, and a table of authorities.
Quick Check and Key Cite Overruling Risk are examples of tools that leverage AI to expand online legal research beyond just searching into deeper analysis.
Stage 6: Analytics for case strategy
Innovation — data-driven legal strategy techniques
Legal data consists of more than just the content of court decisions. Every legal document — case, statute, regulation, motion, pleading, or brief — includes “metadata.” Metadata is data about the data itself. In legal sources, metadata encompasses information about parties, judges, legal counsel, and damage award amounts. It also includes data about jurisdiction, specific courts, and the subject matter of the data. Finally, the metadata provides information about the web of legal authorities cited and cross-referenced across a set of legal documents.
Machines are uniquely capable of ingesting and making sense of metadata in legal documents in a way that humans can’t manage on their own due to the size and scope of the data sources. Because of the sheer amount of raw data that AI can now analyze, the scope of legal research has grown beyond its traditional boundaries.
For instance, Westlaw’s Litigation Analytics can provide insights about the judge in your case:
- How often has this judge sided with a plaintiff in this type of case? In this specific type of motion?
- What kind of arguments have historically resonated with this judge?
- What kinds of authority does the judge typically rely on?
Attorneys can use this data to interpret patterns and use that insight to build the strongest case strategy and respond to clients even faster. They can also use those insights to manage client expectations surrounding the outcome, timeline, and cost.
Knowing how often a judge sides with a plaintiff in a certain type of case or how that judge has viewed certain motions is very powerful — especially when those insights come from hundreds or thousands of data points rather than from a few anecdotes from colleagues or personal experience.
Litigation insights can also track data about opposing counsel or their law firm to obtain insights about their experience on a particular issue, their success at specific types of motions, and whether they have a significant history with a given judge.
All this data gives legal researchers more insights and helps them create their strategies and pleadings with confidence about how the court and opposing counsel might respond.
Stage 7: Case analysis and workflow
Innovation — enhanced analysis, outline generation, and visualization
Earlier stages of legal research innovation focused on improving the process of finding relevant sources and precedents. But even the best search tools leave much work on lawyers' shoulders as they read, filter, review, and organize retrieved documents to align with the case they are working on. The introduction of Westlaw Precision with CoCounsel in 2022 responded to the need to drill down the search results that are most relevant to their current research issue more efficiently.
As with earlier innovations, Westlaw built these innovations on a combination of trusted and enhanced data, exclusive editorial enhancements, and cutting-edge technology. Westlaw Precision with CoCounsel allows you to find what you need quickly without sacrificing confidence in your results.
The following are some of the new capabilities introduced.
Precision Research
This feature builds on the expansion of the human editorial enhancements by Thomson Reuters that improve and expand the data on which Westlaw operates. Where earlier innovations leveraged the classification of cases into the Key Number System, Westlaw Precision with CoCounsel also classifies cases by other attributes such as issue outcome, fact pattern, motion type, motion outcome, cause of action, and party type.
That enhanced data helps users find legally and factually similar cases with unparalleled speed and accuracy. Researchers often look for one case that addresses a particular issue, in a particular context, or with a particular outcome. They may be looking for a specific party type or fact pattern. They may want to identify cases with one distinct cause of action or a precise kind of motion. Ideally, they are looking for material facts that match their case.
Precision Research reveals these critical case attributes as part of the research process, enabling researchers to search, filter, and browse more efficiently than ever before.
As with earlier innovations, Westlaw built these innovations on a combination of trusted and enhanced data, exclusive editorial enhancements, and cutting-edge technology. Westlaw Precision with CoCounsel allows you to find what you need quickly without sacrificing confidence in your results.
The following are some of the new capabilities introduced.
KeyCite Cited With and Overruled features
Westlaw Precision with CoCounsel also introduced new KeyCite capabilities. KeyCite Cited With shows related cases with a pattern of being cited together, even if neither cites the other. It helps legal researchers quickly find connections between cases that would traditionally be difficult to uncover and locate additional relevant authority for the research issue. It also allows researchers to find regularly cited law for a specific rule or issue.
The new KeyCite Overruled in Part flag shows you the specific point of law in a case that has been invalidated so that other valid points of law aren’t missed.
Optimized research workflow
Outline Builder aids in the transition from research to drafting. Researchers can build outlines quickly and easily without needing to leave Westlaw Precision with CoCounsel. Outline Builder can be used alongside a document while researching so important text or citations can be dragged and dropped directly into an outline. Linked and formatted citations and KeyCite flags are automatically included.
Visualization
Graphical View of History displays a visualization of research history, graphically mapping out the steps taken and the searches and documents that were interacted with the most. This feature helps identify the most important documents and keep them front and center. The Keep List/Hide Details capability allows researchers to save cases of interest and hide cases they don’t want to see again.
Stage 8: Generative AI comes to legal research
Innovation — large-language models plus retrieval-augmented generation
Most lawyers are familiar with the excitement around generative AI. For example, they might have tested ChatGPT or Copilot in the Bing search engine. The large-language models (LLMs) behind these GenAI tools are very good at producing a well-written answer to a question that is responsive to the user’s specific questions and sounds accurate and authoritative.
The “sound accurate and authoritative” part concerns many lawyers, however.
The freely available generative AI tools that most people have used have been trained on a wide variety of content captured from the Internet and other sources. Lawyers who have tested those systems with legal questions see many shortcomings. A New York lawyer was recently sanctioned after submitting a court brief that sounded like it contained convincing legal arguments but contained references to non-existent cases presented as legal authority. Anyone who has used the Internet regularly knows it is full of non-authoritative content.
It turns out that popular GenAI systems are good at the “generative” part but not as good at the “authoritative” part. In the generally available tools, there is more emphasis on generating language that sounds good and less emphasis on accuracy in the specifics of how the systems’ answers are sourced or on ensuring that the information returned is up to date.
Fortunately, there are sophisticated techniques that address that problem. Westlaw Precision with CoCounsel has recently incorporated a new capability called AI-Assisted Research. It delivers a solution that lawyers can trust and significantly enhances the speed with which they find the answers to their legal research questions.
How AI-Assisted Research works
AI-Assisted Research gets around the limitations of generative AI in a few ways. First, the tool begins by looking for applicable primary law before generating text. The underlying content searched is essential. Many LLMs use training data that’s only current up to a point in time, and their responses won’t reflect anything that happened after that time. But the law continues to evolve, and because the LLM relies on Westlaw’s primary law, AI-Assisted Research is constantly taking in and learning from the latest and most relevant law.
This process, called research-automated generation (RAG), takes all the AI has learned from a broad knowledge base but applies that knowledge to the more focused data sources identified from the Westlaw database. The selection of the relevant, current, and authoritative sources in this process leverages many of the previous innovations we have discussed above, including the editorial enhancements that Westlaw attorney-editors add to these primary law sources, such as West Key Number System classifications, headnotes, and KeyCite analysis.
Only after it has identified a more targeted set of relevant data and documents does it generate a response to the research question. It then delivers the answer as a well-written response that answers the legal question. Notably, the response also includes footnotes so the researcher can see the legal passages that informed the response.
What it means for legal research
The use of RAG within Westlaw Precision with CoCounsel’s AI-Assisted Research addresses several common concerns expressed by lawyers about relying on AI to help them with legal research:
- Authority. It bases its responses on a set of data culled from the authoritative set of sources contained in Westlaw rather than on the broader world of unverified data that popular generative AI systems like ChatGPT use.
- Currency. It bases its responses on up-to-date and verified data, unlike more general large-language models limited to data existing at some defined date in the past.
- Verifiability. It puts more control in the hands of the lawyer. Unlike the “black box” responses that other systems provide, AI-Assisted Research delivers a complete answer to the research question but also identifies the specific sources relied on for that answer so lawyers can verify the accuracy of the results .
This combination of GenAI and existing Westlaw data and capabilities brings new dimensions to legal research.
“The ability to type a question, get an answer, and have all the supporting resources right underneath that answer so you can ensure it is supported by case law within Westlaw is very important. And I think it’s proven to be a very effective and accurate way to get answers to complex legal questions,” said Andrew Bedigian, Counsel, Larson LLP.
He added, “As lawyers, we should be cautious when using AI to develop responses to complex legal issues that are often dependent on nuance. Because AI-Assisted Research relies on Thomson Reuters’ proven database, I can have confidence that a response generated from AI relies on actual sources and not something that is made up.”
His other big piece of advice? “No tool can do all of your research for you. No matter what you’re using to understand the law, be sure you’re checking the sources the results come from and using it as a starting point for your research. The real wisdom still comes from the legal professional, not the research system.”
Technology for finding the right answer, not just the right documents
AI-Assisted Research is not a stand-alone AI technology that Thomson Reuters has simply bolted onto Westlaw Precision with CoCounsel. As this look back at the history of innovation in legal research shows, several previous generations of research innovations shaped a generative AI solution uniquely suited to today’s lawyers and legal researchers.
Generative AI and many of these other innovations have combined to help lawyers arrive at comprehensive, relevant answers more quickly and efficiently. Most importantly, they do so in a way that leaves the lawyer in charge of the outcome, with a level of transparency that helps lawyers verify that answers are correct and on point to serve clients' needs better.
Get a jumpstart on your research with faster answers to legal questions with Westlaw Precision with CoCounsel.
