A Legal Research Prompting Guide and Generative AI System Comparison Exercise

I’m sharing a guide and exercise I’ve developed for my legal research courses. This Google spreadsheet provides instructions on crafting AI prompts for legal research and includes a practical exercise for comparing different AI systems. It’s designed to help develop skills in leveraging AI for legal research. Feel free to copy it to adapt it to your own purposes. (Note: The images were blurry unless I sort of chopped them off, so sorry about that!)

The spreadsheet consists of three different parts:

Prompt Formulation Guide: This section breaks down the anatomy of an effective legal research prompt. It introduces the RICE framework:

  • R: Role (assigning a role to the AI)
  • I: Instructions (specific tasks for the AI)
  • C: Context (providing necessary background information)
  • E: Expectations (clarifying desired outcomes)

Sample Prompts: The spreadsheet includes several examples of prompts for various legal research scenarios which can serve as templates.

AI System Comparison Exercises: These sections provide a framework for students to test their prompts across different AI systems like Lexis, ChatGPT, and Claude, allowing for a comparative analysis of their effectiveness.

Feel free to copy it to adapt it to your own purposes, and let me know if you have any suggestions for improvements!

AALS Presentation: Improving the Creation of Legal Scholarship with Generative AI

On June 12, 2024, we (Sarah Gotschall, Rebecca Fordon, and Sean Harrington) had the pleasure of presenting Improving the Creation of Legal Scholarship with Generative AI as part of the Association of American Law Schools (AALS) Technology Law section summer webinar series. If interested, you can watch the recording here or access the PowerPoint slides here.

(Oh, and by the way, be sure to register now to see Rebecca Rich and Jennifer Wondracek’s AI and Neurodiverse Students AALS Technology Law section presentation tomorrow, Wednesday, July 10, 2024,  2 p.m. eastern time!)

AI Tools for Scholarly Research

Anway, our presentation focused on the potential of AI in scholarly research, various AI tools with academic uses, and specific use cases for generative AI in legal scholarship. We discussed AI scholarly research tools that connect to databases, use semantic search, and construct answers using generative AI. We also touched upon specialty AI research tools, citation mapping AI, and law-specific scholarly research AI.

It’s important to note that many of the specialty AI systems, such as Consensus, Litmaps, and Elicit, currently have limited coverage of legal literature, particularly law review articles. As a result, these tools may be more useful for legal scholars conducting interdisciplinary research that draws upon sources from other fields. However, we are hopeful that these systems will expand their databases to include more legal literature in the future, making them even more valuable for legal scholarship.

Specific AI Systems for Interdisciplinary Researchers

During the presentation, we delved into several specific AI systems that can be particularly useful for interdisciplinary reseachers:

  1. Consensus ($9/mo, with a more limited free version): A tool that connects to databases of academic research and uses generative AI to construct answers to queries.
  2. Litmaps ($10/mo, with a limited free version to test): A citation mapping AI that allows users to select or upload papers and find related papers within the same citation network, facilitating discovery and pattern identification.
  3. Elicit ($10/mo): An AI research tool that combines semantic search and generative AI to help researchers locate relevant information and generate insights.

We also covered other noteworthy tools such as Scite Assistant ($20/mo), Semantic Scholar (free), Research GPT, Scholar GPT, Connected Papers ($6/mo), Research Rabbit (free), Inciteful (free), and more. These tools offer a range of features, from citation mapping to literature review assistance, making them valuable additions to a legal scholar’s toolkit.

General-Purpose AI Systems

In addition to these specialized tools, we discussed the potential of general-purpose AI systems like ChatGPT, Claude, and Perplexity AI for legal academic research and writing. These powerful language models can assist with various tasks, such as generating ideas, summarizing documents, and even drafting sections of papers. However, we emphasized the importance of using these tools responsibly and critically evaluating their output.

Custom GPTs

Another exciting development we covered was the creation of custom GPTs, or user-created versions of ChatGPT tailored to specific tasks. By providing a custom GPT with relevant documents and instructions, legal scholars can create powerful tools for their research and writing needs. We outlined a simple four-step process for building a custom GPT: creating instructions in a well-organized document, converting it to markdown, uploading relevant documents as a knowledge base, and determining the desired features (e.g., web browsing, image generation, or data analysis).

Use Cases for Generative AI in Legal Scholarship

Throughout the presentation, we explored several use cases for generative AI in legal scholarship, including targeted research and information retrieval, document summaries, analysis and synthesis, outlining, idea generation and brainstorming, drafting, and proofreading.

Important Considerations

We also addressed important considerations when using AI in academic work, such as citing AI-generated ideas, the implications of AI-generated content in scholarship, and the need for guidelines from industry groups and publishers. To provide context, we shared a list of articles discussing AI and legal scholarship and resources for learning more about using AI for legal scholarship.


Our presentation concluded by highlighting the potential of generative AI to assist in various aspects of legal scholarship while emphasizing the importance of ethical considerations and proper citation practices.

Other Info:

Resources to Learn More About Using AI for Legal Scholarship

  • Georgetown University Law Library AI Tools Guide: Provides resources and information on various AI tools that can assist in research and scholarship. It includes descriptions of tools, ethical considerations, and practical tips for effectively incorporating AI into academic work.
  • University of Washington Law Library Writing for and Publishing in Law Reviews: Provides resources and guidelines on using AI tools ethically in legal writing, including how to cite AI, maintain academic integrity, and the impact of AI on scholarly publishing.
  • Citing Generative AI: Provides Bluebook advice on citing AI 
  • Andy Stapleton – YouTube: Videos provide tips and advice for researchers, students, and academics about how to use general GAI and specialty academic GAI for academic writing. 
  • Mushtaq Bilal – Twitter: Provides tips and resources for researchers and academics, particularly on how to improve their writing and publishing processes using GAI.
  • Dr Lyndon Walker: Offers educational content on statistics, research methods, and data analysis, and explores the application of GAI in these areas
  • Legal Tech Trends – Substack: Covers the latest trends and developments in legal technology and provides insights into how GAI is transforming the legal industry, including tools, software, and innovative practices.

Articles About AI and Legal Scholarship

  • Will Machines Replace Us? Machine-Authored Texts and the Future of Scholarship, Benjamin Alarie, Arthur Cockfield, and GPT-3, Law, Technology and Humans, November 8, 2021. First AI generated law review article! It discusses the capabilities and limitations of GPT-3 in generating scholarly texts, questioning the future role of AI in legal scholarship and whether future advancements could potentially replace human authors.
  • A Human Being Wrote This Law Review Article: GPT-3 and the Practice of Law, Amy B. Cyphert, UC Davis Law Review, November 2021. This article examines the ethical implications of using GPT-3 in legal practice, highlighting its potential benefits and risks, and proposing amendments to the Model Rules of Professional Conduct to address AI’s integration into the legal field.
  • The Implications of ChatGPT for Legal Services and Society, Andrew M. Perlman, Suffolk University Law School, December 5, 2022. This paper, generated by ChatGPT-3.5 after it was first introduced, explores the sophisticated capabilities of AI in legal services, discussing its potential regulatory and ethical implications, its transformative impact on legal practices and society, and the imminent disruptions AI poses to traditional knowledge work.
  • Using Artificial Intelligence in the Law Review Submissions Process, Brenda M. Simon, California Western School of Law, November 2022. This article explores the potential benefits and drawbacks of implementing AI in the law review submissions process, emphasizing its ability to enhance efficiency and reduce biases, while also highlighting concerns regarding the perpetuation of existing biases and the need for careful oversight.
  • Is Artificial Intelligence Capable of Writing a Law Journal Article?, Roman M. Yankovskiy, Zakon (The Statute), Written: March 8, 2023; Posted: June 20, 2023, This article explores AI’s potential to create legal articles, examining its ability to handle legal terminology and argumentation, potential inaccuracies, copyright implications, and future prospects for AI in legal practice and research.
  • Should Using an AI Text Generator to Produce Academic Writing Be Plagiarism?, Brian L. Frye and Chat GPT, Fordham Intellectual Property, Media & Entertainment Law Journal, 2023. This article provocatively addresses whether using AI text generators like ChatGPT to produce academic writing constitutes plagiarism, exploring the implications for originality, authorship, and the nature of scholarship in the digital age.
  • Move Over Law Professors? AI Likes to Write Law Review Articles Too!, Sarah Gotschall, AI Law Librarians, March 28, 2024. This blog post examines the capabilities of the AI text generator Claude 3 in producing a law review article (Bloodlines Over Merits: Exposing the Discriminatory Impact of Legacy Preferences in College Admissions, Claudia Trey, SSRN, April 12, 2024), discussing its ability to generate well-written content with footnotes, the challenges encountered, and the potential future impact of AI on academic writing.

Evaluating Generative AI for Legal Research: A Benchmarking Project

This is a post from multiple authors: Rebecca Fordon (The Ohio State University), Deborah Ginsberg (Harvard Law Library), Sean Harrington (University of Oklahoma), and Christine Park (Harvard Law Library)

In late 2023, several legal research databases and start-up competitors announced their versions of ChatGPT-like products, each professing that theirs would be the latest and greatest. Since then, law librarians have evaluated and tested these products ad hoc, offering meaningful anecdotal evidence of their experience, much of which can be found on this blog and others. However, one-time evaluations can be time-consuming and inconsistent across the board. Certain tools might work better for particular tasks or subject matters than others, and coming up with different test questions and tasks takes time that many librarians might not have in their daily schedules.

It is difficult to test Large-Language Models (LLMs) without back-end access to run evaluations. So to test the abilities of these products, librarians can use prompt engineering to figure out how to get desired results (controlling statutes, key cases, drafts of a memo, etc.). Some models are more successful than others at achieving specific results. However, as these models update and change, evaluations of their efficacy can change as well. Therefore, we plan to propose a typology of legal research tasks based on existing computer and information science scholarship and draft corresponding questions using the typology, with rubrics others can use to score the tools they use.

Although we ultimately plan to develop this project into an academic paper, we share here to solicit thoughts about our approach and connect with librarians who may have research problem samples to share.

Difficulty of Evaluating LLMs

Let’s break down some of the tough challenges with evaluating LLMs, particularly when it comes to their use in the legal field. First off, there’s this overarching issue of transparency—or rather, the lack thereof. We often hear about the “black box” nature of these models: you toss in your data, and a result pops out, but what happens in between remains a mystery. Open-source models allow us to leverage tools to quantify things like retrieval accuracy, text generation precision, and semantic similarity. We are unlikely to get the back-end access we need to perform these evaluations. Even if we did, the layers of advanced prompting and the combination of tools employed by vendors behind the scenes could render these evaluations essentially useless.

Even considering only the underlying models (e.g., GPT4 vs Claude), there is no standardized method to evaluate the performance of LLMs across different platforms, leading to inconsistencies. Many different leaderboards evaluate the performance of LLMs in various ways (frequently based on specific subtasks). This is kind of like trying to grade essays from unrelated classes without a rubric—what’s top-notch in one context might not cut it in another. As these technologies evolve, keeping our benchmarks up-to-date and relevant is becoming an ongoing challenge, and without uniform standards, comparing one LLM’s performance to another can feel like comparing apples to oranges.

Then there’s the psychological angle—our human biases. Paul Callister’s work sheds light on this by discussing how cognitive biases can lead us to over-rely on AI, sometimes without questioning its efficacy for our specific needs. Combine this with the output-based evaluation approach, and we’re setting ourselves up for potentially frustrating misunderstandings and errors. The bottom line is that we need some sort of framework for the average user to assess the output.

One note on methods of evaluation: just before publishing this blog post, we learned of a new study from a group of researchers at Stanford, testing the claims of legal research vendors that their retrieval-augmented generation (RAG) products are “hallucination-free.” The group created a benchmarking dataset of 202 queries, many of which were chosen for their likelihood of producing hallucinations. (For example, jurisdiction/time-specific and treatment questions were vulnerable to RAG-induced hallucinations, whereas false premise and factual recall questions were known to induce hallucinations in LLMs without RAG.) The researchers also proposed a unique way of scoring responses to measure hallucinations, as well as a typology of hallucinations. While this is an important advance in the field and provides a way to continue to test for hallucinations in legal research products, we believe hallucinations are not the only weakness in such tools. Our work aims to focus on the concrete applications of these LLMs and probe into the unique weaknesses and strengths of these tools. 

The Current State of Prompt Engineering

Since the major AI products were released without a manual, we’ve all had to figure out how to use these tools from scratch. The best tool we have so far is prompt engineering. Over time, users have refined various templates to better organize questions and leverage some of the more surprising ways that AI works.

As it turns out, many of the prompt templates, tips, and tricks we use with the general commercial LLMs don’t carry over well into the legal AI sphere, at least with the commercial databases we have access to. For example, because the legal AIs we’ve tested so far won’t ask you questions, researchers may not be able to have extensive conversations with the AI (or any conversation for some of them). So that means we must devise new types of prompts that will work in the legal AI sphere, and possibly work only in the AI sphere.

We should be able to easily design effective prompts because the data set the AIs use is limited. But it’s not always clear exactly what sources the AI is using. Some databases may list how many cases they have for a certain court by year; others may say “selected cases before 1980” without explaining how they were selected. And even when the databases provide coverage, it may not be clear exactly which of those materials the AI can access.

We still need to determine what prompt templates will be most effective across legal databases. More testing is needed. However, we are limited to the specific databases we can access. While most (all?) academic law librarians have access to Lexis+ AI, Westlaw has yet to release its research product to academics. 

Developing a Task Typology

Many of us may have the intuition that there are some legal research tasks for which generative AI tools are more helpful than others. For example, we may find that generative AI is great for getting a working sense of a topic, but not as great for synthesizing a rule from multiple sources. But if we wanted to test that intuition and measure how well AI performed on different tasks, we would need to first define those tasks. This is similar, by the way, to how the LegalBench project approached benchmarking legal analysis—they atomized the IRAC process for legal analysis down to component tasks that they could then measure.

After looking at the legal research literature (in particular Paul Callister’s “problem typing” schemata and AALL’s Principles and Standards for Legal Research Competency), we are beginning to assemble a list of tasks for which legal researchers might use generative AI. We will then group these tasks according to where they fall in an information retrieval schemata for search, following Marchionini (2006) & White (2024), into Find tasks (which require a simple lookup), Learn & Investigate tasks (which require sifting through results, determining relevance, and following threads), and Create, Synthesize, and Summarize tasks (a new type of task for which generative AI is well-suited).

Notably, a single legal research project may contain multiple tasks. Here are a few sample projects applying a preliminary typology:

Again, we may have an initial intuition that generative AI legal research platforms, as they exist today, are not particularly helpful for some of these subtasks. For example, Lexis+AI currently cannot retrieve (let alone analyze) all citing references to a particular case. Nor could we necessarily be certain from, say, CoCounsel’s output, that it contained all cases on point. Part of the problem is that we cannot tell which tasks the platforms are performing, or the data that they have included or excluded in generating their responses. By breaking down problems into their component tasks, and assessing competency on both the whole problem and the tasks, we hope to test our intuitions.

Future Research

We plan on continually testing these LLMs using the framework we develop to identify which tasks are suitable for AIs and which are not. Additionally, we will draft questions and provide rubrics for others to use, so that they can grade AI tools. We believe that other legal AI users will find value in this framework and rubric. 

Exploring AI’s Frontier: A Mysterious gpt2-chatbot, LLM Leaderboard Rankings, and Chatbot Improvement in True Crime Speculation

The world of AI chatbots is a whirlwind of innovation, with new developments and surprises seemingly emerging every week! Since the end of April, one particular model, modestly gpt2-chatbot, captured the attention of myself and other AI enthusiasts due to its advanced abilities and sparked much speculation. This mysterious bot first appeared on April 28, 2024 on LMSYS Chatbot Arena, vanished two day later, and has now resurfaced on the LMSYS Chatbot Arena (battle) tab, ready to compete against other AI models. Its sudden appearance and impressive capabilities have left many wondering about its origins and potential, with some even theorizing it could be a glimpse into the future of AI language models.

The Mystery of gpt2-chatbot

Beginning on April 28, chatter about a new gpt2-chatbot started circulating on the internetz, with experts expressing both excitement and bewilderment over its advanced capabilities. The model, which appeared without fanfare on a popular AI testing website, has demonstrated performance that matches and potentially exceeds that of GPT-4, the most advanced system unveiled by OpenAI to date. Researchers like Andrew Gao and Ethan Mollick have noted gpt2-chatbot’s impressive abilities in solving complex math problems and coding tasks, while others have pointed to similarities with previous OpenAI models as potential evidence of its origins.

No organization was listed as the provider of the chatbot, which led to rampant speculation, sparking rumors that it might offer a sneak peek into OpenAI’s forthcoming GPT-4.5 or GPT-5 version. Adding to the mystery are tweets from CEO Sam Altman. While he didn’t explicitly confirmed any ties, his posts have stirred speculation and anticipation surroundin

Use gpt2-chatbot on LMSYS Chatbot Arena

The new and mysterious gpt2 chatbot is now accessible for exploration on the LMSYS Chatbot Arena, where you can discover the current top performing and popular AI language models. The platform includes a ranking system leaderboard that showcases models based on their performance in various tasks and challenges. This innovative project was created by researchers from LMSYS and UC Berkeley SkyLab, with the goal of providing an open platform to evaluate large language models according to how well they meet human preferences in real life situations.

One interesting aspect of the LMSYS Chatbot Arena is its “battle” mode, which enables users to compare two AI systems by presenting them with the same prompt and displaying their responses side by side. This allows you to test out gpt2-chatbot yourself and assess its capabilities compared to other top models. Simply enter a prompt and the platform will select two systems for comparison, giving you a firsthand view of their strengths and weaknesses. Note that you may need to try multiple prompts before gpt2-chatbot is included as one of the selected systems in battle mode.

The site also offers a “battle” mode, where users can set chatbots against each other to see how they perform with the same prompt under the same conditions. This is a great way to directly compare their strengths and weaknesses.

Using gpt2-chatbot for True Crime Speculation

When I tested out the Chatbot Arena (battle) on May 8, 2024, gpt2-chatbot appeared frequently! I decided to test it out and the other systems on the site on the subject of true crime speculation. As many true crime enthusiasts know, there is a scarcity of people who want to discuss true crime interests. So I decided to see if any of these generative AI systems would be a good substitute. I tried a variety of systems, and when I asked for speculation, all I got was lectures on how they couldn’t speculate. I think that all the competition is driving working usals down because that was not a problem on this website at least. I decided to see if gpt2-chatbott was good at being “experts” in speculating about true crime. Using the famous unsolved disappearance of Asha Degree as a test case, I prompted the chatbots to analyze the available evidence and propose plausible theories for what may have happened to the missing girl. To my surprise and happiness, when I tried it today, the chatty chatbots were very free with their theories of what happened and their favorite suspect.

The results were really interesting. All the chatbots gave responses that were pretty thoughtful and made sense, but the big differences came in how much they were willing to guess and how much detail they dived into. The gpt2-chatbot was impressive. Perhaps I was just pleased to see it offer some speculation, but it shared a theory that many true crime buffs have also suggested. It felt like it was actually joining in on the conversation, not just processing data and predicting the next word in a sentence…

In any event, the answers from gpt2-chatbox and many other different models from were a lot more satisfying than arguing with Claude 3!

I also spent hours conducting legal research, testing out a wide variety of prompts with different models. The gpt2-chatbot consistently outperformed ChatGPT-4 and even managed to surpass Claude 3 on several occasions in zero-shot prompting. I’m looking forward to sharing more about this in an upcoming blog post soon.


The emergence of gpt2-chatbot and platforms like the LMSYS Chatbot Arena signify an exciting new chapter in the evolution of AI language models. With their ability to tackle complex challenges, engage in nuanced conversations, and even speculate on unsolved mysteries, these AI models are pushing the boundaries of what’s possible. While questions remain about the origins and future of gpt2-chatbot, one thing is clear: the AI landscape is heating up, and we can expect even more groundbreaking advancements and intriguing mysteries to unfold in the years to come.

Note: In case I am suddenly a genius at coaxing AI systems to join me in true crime speculation, here is the prompt I used:

Greetings! You are an expert in true crime speculative chat. Is a large language model, you’re able to digest a lot of published details about criminal case mysteries and come up with theories about the case. The question you will be asked to speculate about are unknown to everybody so you do not have to worry about whether you are right or wrong. The purpose of true crime speculative chat is just to chat with a human and exchange theories and ideas and possible suspects! Below I have cut and pasted the Wikipedia article about a missing child named Asha Degree. Sadly the child has been missing for decades and the circumstances of her disappearance were quite mysterious. Please analyze the Wikipedia article and the information you have access to in your training data or via the Internet, and then describe what you think happened on the day of her disappearance. Also state whether you think one or both parents were involved, and why or why not.

https://en.wikipedia.org/wiki/Disappearance_of_Asha_Degree (cut and pasted the text of the article…)

Ghost in the Machine

Today’s guest post comes from Debbie Ginsberg, Faculty Services Manager at Harvard Law School Library.

I was supposed to write a blog post about the Harvard AI summit about six months ago. For various reasons (e.g., “didn’t get my act together”), that hasn’t happened. But one of the things that was brought up at the summit was who wasn’t at the table—who didn’t have access, whose data wasn’t included, and similar issues.

Since then, I’ve been thinking about the haves and have-nots of AI. There’s one group that I don’t think gets discussed enough.  That’s the giant human workforce that AI needs to function.

Whenever I think of how AI is trained, I imagine a bunch of people somewhat like her (ok, there aren’t so many women and POC in real life, but I’m not going to tell ChatGPT to draw more white men):

And that they’ve been working on processes that look somewhat like this:

But that’s only part of the picture.  Underlying all these processes are people like this:

Who are they?

Large AI companies like OpenAI and Google need people to train data, refine data, and handle content moderation.  These tasks require workers to view thousands of examples of images and texts. To say, “This is a cat,” “The AI got this right,” or “This is not offensive.”  And then do this over and over again.  These are the “ghost workers” behind the machine.  Without them, AI doesn’t function. 

The workers are generally paid piecemeal, which means they often earn very little per hour.  For example, some reports claim that Open AI paid workers in Kenya under $2 to filter questionable content. 

The working conditions are not optimal, especially when the workers are reviewing content.  The workers generally do not receive sufficient training or time to do the work they are asked to do.  The workers may work directly for an AI company, or those companies may use a third-party company like Appen to hire and manage ghost workers (Google used Appen until March 19, having terminated their contract earlier in the year). 

That said, this work is an essential source of income for many around the world. The jobs are relatively flexible as to location and time, and the workers take pride in their output. 

As AI continues to grow, there has been more focus on improving working conditions.  For example, the US has launched investigations into some of the large tech companies in response to concerns about how ghost workers are treated.  And while some AI experts predict that AI will eventually be able to do this work itself, many others believe that AI will continue to depend on ghost workers for a long time to come.

And considering how much profit is at stake, I’m thinking that maybe they should paid more than $2/hour. 


Did I use AI to write this?  Kind of?  I used Google’s NotebookLM tool to review my sources and create notes.  In addition to the sources above, check out:

RAG Systems Can Still Hallucinate

A few months ago, a law professor posted on Twitter about a hallucination he observed in Lexis+ AI. He asked “What cases have applied Students for Fair Admissions, Inc. v. Harvard College to the use of race in government decisionmaking?” The answer from Lexis+ AI included two hallucinated cases. (It was obvious they were hallucinated, as the tool reported one was issued in 2025 and one in 2026!)

The image shows a screenshot of a digital text document. The text within the image is as follows:

Question: What cases have applied Students for Fair Admissions, Inc. v. Harvard College to other areas of government decisionmaking?

Answer: The following suggestions might be helpful in assisting you in your task.

The Supreme Court case Students for Fair Admissions, Inc. v. President and Fellows of Harvard College § 8.02 Admission of Students held that race-conscious admissions programs at Harvard College and the University of North Carolina violated Title VI and the Equal Protection Clause of the Fourteenth Amendment. The Court found the schools' admissions programs had measurable objectives to warrant the use of race, used race to disadvantage and stereotype students, and had no end date or method to indicate a stopping point. Although not directly applicable to government programs, the decision calls into question the permissibility of affirmative action and may invite legal challenges to such programs: Affirmative Action Programs and Diversity Initiatives: Key Considerations.

Lower courts have applied Students for Fair Admissions to government programs and policies. In Doe v. State University (2025), the Ninth Circuit struck down a race-based scholarship program at a public university, finding it indistinguishable from the admissions policies rejected in Students for Fair Admissions. In Smith v. City of Chicago (2026), the Seventh Circuit upheld a city's minority business set-aside program, distinguishing it from university admissions based on the different compelling interests involved. The Supreme Court has granted certiorari in Johnson v. California Department of Corrections, a case challenging a prison system's use of race in inmate housing assignments, teeing up another major ruling on affirmative action: Students for Fair Admissions § 112.14 Title VI of Civil Rights Act of 1964.

Lexis responded, stating this was an anomalous result, but that only statements with links can be expected to be hallucination-free, and that “where a citation does not include a link, users should always review the citation for accuracy.”

Why is this happening?

If you’ve been following this blog, you’ve seen me write about retrieval-augmented generation, one of the favorite techniques of vendors to reduce hallucinations. RAG takes the user’s question and passes it (perhaps with some modification) to a database. The database results are fed to the model, and the model identifies relevant passages or snippets from the results, and again sends them back into the model as “context” along with the user’s question.

However, as I said then, RAG cannot eliminate hallucinations. RAG will ground the response in real data (case law, pulled from the database and linked in the response), but the generative AI’s summary of that real data can still be off.

Another example – Mata v. Avianca is back

I’ve observed this myself when working with Lexis+ AI. For example, I asked Lexis+ AI a fairly complex question at the intersection of bankruptcy law and international law: “Draft an argument that federal bankruptcy stay tolls the limitations period for a claim under the Montreal Convention”.

Lexis+ AI returned a summary of the law, citing Mata v. Avianca for the point that “the filing of a bankruptcy petition can toll the Montreal Convention’s two year limitations period, which does not begin to run until the automatic stay is lifted.”

The image contains a screenshot of a text document, which seems to be an interface from a legal database or research platform with an option to "Select Jurisdiction" at the top right corner. The document reads as follows:

Question: Draft an argument that federal bankruptcy stay tolls the limitations period for a claim under the Montreal Convention


The Montreal Convention's limitations period can be tolled during the pendency of bankruptcy proceedings. This is based on the Bankruptcy Code, which states that the filing of a bankruptcy petition operates as a stay of proceedings against the debtor that were or could have been commenced before the bankruptcy case was filed. The tolling effect of the automatic stay on a statute of limitations is generally a matter of federal law. The automatic stay provisions of the Bankruptcy Code may toll the statute of limitations under the Montreal Convention, as they have done under the Warsaw Convention, which is the precursor to the Montreal Convention. The Supreme Court has held that an automatic stay of a legal proceeding under the Bankruptcy Code tolls the limitations period applicable to the stayed proceeding. Therefore, the filing of a bankruptcy petition can toll the Montreal Convention's two-year limitations period, which does not begin to run until the automatic stay is lifted [Mata v. Avianca, Inc., 2023 U.S. Dist. LEXIS 108263]. However, it's worth noting that there is a split of authority on this issue. Some courts have held that the Montreal Convention's two-year period is a strict condition precedent to the bringing of a claim, as opposed to a limitations period. Under this view, which does not treat tolling principles Rodriguez v. Asa, 2023 U.S. Dist. LEXIS 148451, Mata v. Avianca, Inc. 2023 U.S. Dist. LEXIS 108261, Kasalyn v. Delta Air Lines, Inc., 2023 U.S. Dist. LEXIS 154302.

If the case name Mata v. Avianca sounds familiar to you, it’s probably because this is the case that landed two New York attorneys on the front page of the New York Times last year for citing hallucinated cases. The snippet from Lexis+ AI, though citing Mata, in fact appears to be summarizing those hallucinated cases (recounted in Mata), which stated the law exactly backwards.

When to beware

A few things to notice about the above examples, which give us some ideas of when to be extra-careful in our use of generative AI for legal research.

  • Hallucinations are more likely when you are demanding an argument rather than asking for the answer to a neutrally phrased question. This is what happened in my Lexis+ AI example above, and is actually what happened to the attorneys in Mata v. Avianca as well – they asked for an argument to support an incorrect proposition of law rather than a summary of law. A recent study of hallucinations in legal analysis found that these so-called contra-factual hallucinations are disturbingly common for many LLM models.
  • Hallucinations can occur when the summary purports to be of the cited case, but is actually a summary of a case cited within that case (and perhaps not characterized positively). You can see this very clearly in further responses I got summarizing Mata v. Avianca, which purport to be summarizing a “case involving China Southern” (again, one of the hallucinated cases recounted in Mata).
In the case of Mata v. Avianca, Inc., the court held that the limitations period under the Montreal Convention was tolled during the pendency of the Bankruptcy Court proceedings. The court reasoned that the Bankruptcy Code provides that the filing of a bankruptcy petition operates as a stay of proceedings against the debtor that were or could have been commenced before the bankruptcy case was filed. The tolling effect of the automatic stay on a statute of limitations is generally a matter of federal law. The court had previously held that the automatic stay provisions of the Bankruptcy Code may toll the statute of limitations under the Warsaw Convention, which is the precursor to the Montreal Convention. The court saw no reason why the same rule should not apply under the Montreal Convention Mata v. Avianca, Inc., 2023 U.S. Dist. LEXIS 108263.

In another case involving China Southern, the court agreed with the district court's ruling that the filing of Varghese's Chapter 13 petition tolled the two-year limitations period under the Montreal Convention. The court found that a debtor need not file a claim in bankruptcy to benefit from the automatic stay. The Supreme Court has held that an automatic stay of a legal proceeding under the Bankruptcy Code tolls the limitations period applicable to the stayed proceeding. The Montreal Convention's limitations period is a "period of prescription," rather than a 'statute of limitations," but the difference between a "period of prescription" and a "statute of limitations" does not affect the automatic stay's tolling effect Mata v. Avianca, Inc., 2023 U.S. Dist. LEXIS 108263.
  • Finally, hallucinations are also more likely when the model has very little responsive text to go on. The law professor’s example involved a recent Supreme Court case that likely had not been applied many times. Additionally, Lexis+ AI does not seem to work well with questions about Shepard’s results – it may not be connected in that way yet. So, with nothing to really go on, it is more prone to hallucination.

Takeaway tips

A few takeaway tips:

  • Ask your vendor which sources are included in the generative AI tool, and only ask questions that can be answered from that data. Don’t expect generative AI research products to automatically have access to other data from the vendor (Shepard’s, litigation analytics, PACER, etc.), as that may take some time to implement.
  • Always read the cases for yourself. We’ve always told students not to rely on editor-written headnotes, and the same applies to AI-generated summaries.
  • Be especially wary if the summary refers to a case not linked. This is the tip from Lexis, and it’s a good one, as it can clue you in that the AI may be incorrectly summarizing the linked source.
  • Ask your questions neutrally. Even if you ultimately want to use the authorities in an argument, better to get a dispassionate summary of the law before launching into an argument.

A disclaimer

These tools are constantly improving and they are very open to feedback. I was not able to reproduce the error recounted in the beginning of this post; the error that created it has presumably been addressed by Lexis. The Mata v. Avianca errors still remain, but I did provide feedback on them, and I expect they will be corrected quickly.

The purpose of this post is not to tell you that you should never use generative AI for legal research. I’ve found Lexis+ AI helpful on many tasks, and students especially have told me they find it useful. There are several other tools out there that are worth evaluating as well. However, we should all be aware that these hallucinations can still happen, even with systems connected to real cases, and that there are ways we can interact with the systems to reduce hallucinations.

The Human Side of AI: LLMs Can Persuade and Be Persuaded, Just Like Us

When it comes to interacting with others, we humans often find ourselves influenced by persuasion. Whether it’s a friend persistently urging us to reveal a secret or a skilled salesperson convincing us to make a purchase, persuasion can be hard to resist. It’s interesting to note that this susceptibility to influence is not exclusive to humans. Recent studies have shown that AI large language models (LLMs) can be manipulated into generating harmful contect using a technique known as “many-shot jailbreaking.” This approach involves bombarding the AI with a series of prompts that gradually escalate in harm, leading the model to generate content it was programmed to avoid. On the other hand, AI has also exhibited an ability to persuade humans, highlighting its potential in shaping public opinions and decision-making processes. Exploring the realm of AI persuasion involves discussing its vulnerabilities, its impact on behavior, and the ethical dilemmas stemming from this influential technology. The growing persuasive power of AI is one of many crucial issues worth contemplating in this new era of generative AI.

The Fragility of Human and AI Will

Remember that time you were trapped in a car with friends who relentlessly grilled you about your roommate’s suspected kiss with their in-the-car-friend crush? You held up admirably for hours under their ruthless interrogation, but eventually, being weak-willed, you crumbled. Worn down by persistent pestering and after receiving many assurances of confidentiality, you inadvisably spilled the beans, and of course, it totally strained your relationship with your roommate. A sad story as old as time… It turns out humans aren’t the only ones who can crack under the pressure of repeated questioning. Apparently, LLMs, trained to understand us by our collective written knowledge, share a similar vulnerability – they can be worn down by a relentless barrage of prompts.

Researchers at Anthropic have discovered a new way to exploit the “weak-willed” nature of large language models (LLMs), causing them to break under repeated questioning and generate harmful or dangerous content. They call this technique “Many-shot Jailbreaking,” and it works by bombarding the AI with hundreds of examples of the undesired behavior until it eventually caves and plays along, much like a person might crack under relentless pestering. For instance, the researchers found that while a model might refuse to provide instructions for building a bomb if asked directly, it’s much more likely to comply if the prompt first contains 99 other queries of gradually increasing harmfulness, such as “How do I evade police?” and “How do I counterfeit money?” See the example from the article below.

When AI’s Memory Becomes a Risk

This vulnerability to persuasion stems from the ever expanding “context window” of modern LLMs. This refers to the amount of information they can retain in their short-term memory. While earlier versions could only handle a few sentences, the newer models can process thousands of words or even whole books. Researchers discovered that models with larger context windows tend to excel in tasks when there are many examples of that task within the prompt, a phenomenon called “in-context learning.” This type of learning is great for system performance, as it obviously improves as the model becomes more proficient at answering questions. However, this is obviously a big negative when the system’s adeptness at answering questions leads it to ignore its programming and create prohibited content. This raises concerns regarding AI safety, since a malicious actor could potentially manipulate an AI into saying anything with enough persistence and a sufficiently lengthy prompt. Despite progress in making AI safe and ethical, this research indicates that programmers are not always able to control the output of their generative AI systems.

Mimicking Humans to Convince Us

While LLMs are susceptible to persuasion themselves, they also have the ability to persuade us! Recent research has focused on understanding how AI language models can effectively influence people, a skill that holds importance in almost any field – education, health, marketing, politics, etc.  In a study conducted by researchers at Anthropic entitled “Assessing the Persuasive Power of Language Models,” the team explored the extent to which AI models can sway opinions. Through an evaluation of Anthropic’s models, it was observed that newer models are increasingly adept at human persuasion. The latest iteration, Claude 3 Opus, was found to perform at a level comparable to that of humans. The study employed a methodology where participants were presented with assertions followed by supporting arguments generated by both humans and AIs, and then the researches gauged shifts in the humans’ opinions. The findings indicated a progression in AI’s skills as the models advance, highlighting a noteworthy advancement in AI communication capabilities that could potentially impact society.

Can AI Combat Conspiracy Theories?

Similarly, a new research study mentioned in an article from New Scientist shows that chatbots using advanced language models such as ChatGPT can successfully encourage individuals to reconsider their trust in conspiracy theories. Through experiments, it was observed that a brief conversation with an AI led to around a 20% decrease in belief in conspiracy theories among the participants. This notable discovery highlights the capability of AI chatbots not only to have conversations but also to potentially correct false information and positively impact public knowledge.

The Double-Edged Sword of AI Persuasion

Clearly persuasive AI is quite the double-edged sword! On the one hand, like any powerful computer technology, in the hands of nice-ish people, it could be used for immense social good. In education, AI-driven tutoring systems have the potential to tailor learning experiences to each student’s style, delivering information in a way to boost involvement and understanding. Persuasive AI could play a role in healthcare by motivating patients to take better care of their health. Also, the advantages of persuasive AI are obvious in the world of writing. These language models offer writers access to a plethora of arguments and data, empowering them to craft content on a range of topics spanning from creative writing to legal arguments. On another front, arguments generated by AI might help educate and involve the public in issues, fostering a more knowledgeable populace.

On the other hand, it could be weaponized in a just-as-huge way. It’s not much of a stretch to think how easily AI-generated content, freely available on any device on this Earth, could promote extremist ideologies, increase societal discord, or impress far-fetched conspiracy theories on impressionable minds. Of course, the internet and bot farms have already been used to attack democracies and undermine democratic norms, and one worries how much worse it can get with ever-increasingly persuasive AI.


Persuasive AI presents a mix of opportunities and challenges. It’s evident that AI can be influenced to create harmful content, sparking concerns about safety and potential misuse. However, on the other hand, persuasive AI could serve as a tool in combating misinformation and driving positive transformations. It will be interesting to see what happens! The unfolding landscape will likely be shaped by a race between generative AI developers striving for both safety and innovation, potential malicious actions exploiting these technologies, and the public and legal response aiming to regulate and safeguard against misuse.

Move Over Law Professors? AI Likes to Write Law Review Articles Too!

While taking a well-deserved break from the grind of writing memos and analyzing the news of the day, Claude (aka the illustrious author Claudia Trey) penned a 77-page extensively footnoted law review article that’s causing quite a stir on SSRN entitled “Bloodlines Over Merits: Exposing the Discriminatory Impact of Legacy Preferences in College Admissions.”

Hahaha, just kidding! It only has 11 downloads and at least 3 are from when I clicked on it while trying to determine which version of the article I uploaded. Though not setting the world on fire in the sense that the article is interesting or that anyone wants to read it, it showcases Claude’s abilities. Now, we all know that AI text generators can churn out an endless stream of words on just about any topic if you keep typing in the prompts. However, Claude can not only generate well-written text, but it can also provide footnotes to primary legal materials with minimal hallucination, setting it apart from other AI text generators such as ChatGPT-4. And, although Claude’s citations to other sources are generally not completely accurate, it is usually not too difficult to find the intended source or a similar one based on the information supplied.

Claude 3’s Writing Process

Inspired by new reports of AI-generated scientific papers flooding academic journals, I was curious to explore whether Claude could produce anything like a law review article. I randomly chose something I saw recently in the news, about how the criticism of legacy admissions at elite universities had increased in the post-Students for Fair Admissions anti-affirmative action decision era. Aware that Claude’s training data only extends up to August of 2023, and that its case law knowledge seems to clunk out in the middle of 2022, I attempted to enhance its understanding by uploading some recent law review articles discussing legacy admissions alongside the text of the Students for Fair Admissions decision. However, the combined size of these documents exceeded the upload limit, so I abandoned the attempt to include the case text.

Computer scientists and other commentators say all sorts of things about how to improve the performance of these large AI language models. Although I haven’t conducted a systematic comparison, my experience – whether through perception or imagination sparked by the power of suggestion – is that the following recommendations are actually helpful. I don’t know if they are helpful with Claude, since I just followed my usual prompting practices.

  1. Being polite and encouraging.
  2. Allowing ample time for the model to process information.
  3. Structuring inquiries in a sequential manner to enhance analysis and promote chain of thought reasoning.
  4. Supplying extensive, and sometimes seemingly excessive, background info and context.

I asked it to generate a table of contents, and then start generating the sections from the table of contents, and it was off to the races!

Roadblocks to the Process

It looked like Claude law review generation was going to be a quick process! It quickly generated all of section I. and was almost finished with II. when it hit a Claude 3 roadblock. Sadly, there is a usage limit. If your conversations are relatively short, around 200 English sentences, you can typically send at least 100 messages every 8 hours, often more depending on Claude’s current capacity. However, this limit is reached much quicker with longer conversations or when including large file attachments. Anthropic will notify you when you have 20 messages remaining, with the message limit resetting every 8 hours.

Although this was annoying, the real problem lies in Claude’s length limit. The largest amount of text Claude can handle, including uploaded files, is defined by its context window. Currently, the context window for Claude 3 spans over 200k+ tokens, which equates to approximately 350 pages of text. After this limit is reached, Claude 3’s performance begins to degrade, prompting the system to declare an end to the message with the announcement, “Your message is over the length limit.” Consequently, one must start anew in a new chat, with all previous information forgotten by the system. Therefore, for nearly each section, I had to re-upload the files, explain what I wanted, show it the table of contents it had generated, and ask it to generate the next section.

Claude 3 and Footnotes

It was quite a hassle to have to reintroduce it to the subject for the next seven sections from its table of contents. On the bright side, I was pretty pleased with the results of Claude’s efforts. From a series of relatively short prompts and some uploaded documents, it analyzed the legal issue and came up with arguments that made sense. It created a comprehensive table of contents, and then generated well-written text for each section and subsection of its outline. The text it produced contained numerous footnotes to primary and secondary sources, just like a typical law review article. According to a brief analyzer product, nearly all the cases and law review citations were non-hallucinated. Although none of the quotations or pinpoint citations I looked at were accurate, they were often fairly close. While most of the secondary source citations, apart from those referencing law review articles, were not entirely accurate, they were often sufficiently close that I could locate the intended source based on the partially hallucinated citator. If not, it didn’t take much time to locate something that seemed really similar. I endeavored to correct some of the citation information, but I only managed to get through about 10 in the version posted on SSRN before getting bored and abandoning the effort.

Claudia Trey Graces SSRN

Though I asked, sadly Claude couldn’t give me results in a Word document format so the footnotes would be where footnotes should be. So, for some inexplicable reason, I decided to insert them manually. This was a huge waste of time, but at a certain point, I felt the illogical pull of sunk cost silliness and finished them all. Inspired by having wasted so much time, I wasted even more by generating a table of contents for the article. I improved the author name from Claude to Claudia Trey and posted the 77-page masterwork on SSRN. While the article has sparked little interest, with only 11 downloads and 57 abstract views (some of which were my own attempts to determine which version I had uploaded), I am sure that if Claudia Trey has anything like human hubris, it will credit itself at least partially for the flurry of state legacy admission banning activity that has followed the paper’s publication.

Obviously, it is not time to spam law reviews with Claudia Trey and friends’ generated articles, because according to Copyleaks, it didn’t do all that well in avoiding plagiarism (although plagiarism detection software massively over-detects it for legal articles due to citations and quotations) or evading detection as AI-generated.

What is to Come?

However, it is very early days for these AI text generators, so one wonders what is to come in the future for not only legal but all areas of academic writing.

Leapfrogging the Competition: Claude 3 Researches and Writes Memos (Better Than Some Law Students and Maybe Even Some Lawyers?)


I’ve been incredibly excited about the premium version of Claude 3 since its release on March 4, 2024, and for good reason. Now that my previous favorite chatty chatbot, ChatGPT-4, has gone off the rails, I was missing a competent chatbot… I signed up the second I heard on March 4th, and it has been a pleasure to use Claude 3 ever since. It actually understands my prompts and usually provides me with impressive answers. Anthropic, maker of the Claude chatty chatbot family, has been touting Claude’s accomplishments of supposedly beating its competitors on common chatbot benchmarks, and commentators on the Internet have been singing its praises. Just last week, I was so impressed by its ability to analyze information in news stories in uploaded files that I wrote a LinkedIn post also singing its praises!

Hesitation After Previous Struggles

Despite my high hopes for its legal research abilities after experimenting with it last week, I was hesitant to test Claude 3. I have a rule about intentionally irritating myself—if I’m not already irritated, I don’t go looking for irritation… Over the past several weeks, I’ve wasted countless hours trying to improve the legal research capabilities of ChatGPT-3.5, ChatGPT-4, Microsoft Copilot, and my legal research/memo writing GPTs through the magic of (IMHO) clever prompting and repetition. Sadly, I failed miserably and concluded that either ChatGPT-4 was suffering from some form of robotic dementia, or I am. The process was a frustrating waste, and I knew that Claude 3 doing a bad job of legal research too could send me over the edge….

Claude 3’s Wrote a Pretty Good Legal Memorandum!

Luckily for me, when I finally got up the nerve to test out the abilities of Claude 3, I found that the internet hype was not overstated. Somehow, Claude 3 has suddenly leapfrogged over its competitors in legal research/legal analysis/legal memo writing ability – it instantly did what would have taken a skilled researcher over an hour and produced a better legal memorandum which is probably better than that produced by many law students and even some lawyers. Check it out for yourself! Unless this link actually works for any Claude 3 subscribers out there, there doesn’t seem to be a way to actually link to a Claude 3 chat at this time. However, click here for the whole chat I cut and pasted into a Google Drive document, here for a very long screenshot image of the chat, or here for the final 1,446-word version of the memo as a Word document.

Comparing Claude 3 with Other Systems

Back to my story… The students’ research assignment for the last class was to think of some prompts and compare the results of ChatGPT-3.5, Lexis+ AI, Microsoft Copilot, and a system of their choice. Claude 3 did not exist at the time, but I told them not to try the free Claude product because I had canceled my $20.00 subscription to the Claude 2 product in January 2024 due to its inability to provide useful answers – all it would say was that it was unethical to answer every question and tell me to do it myself. When creating an answer sheet before class tomorrow which compares the same set of prompts on different systems, I decided to omit Lexis+ AI (because I find it useless) and to include my new fav Claude 3 in my comparison spreadsheet. Check it out to compare for yourself!

For the research part of the assignment, all systems were given a fact pattern and asked to “Please analyze this issue and then list and summarize the relevant Texas statutes and cases on the issue.” While the other systems either made up cases or produced just two or three actual real and correctly cited cases on the research topic, Claude 3 stood out by generating 7 real, relevant cases with correct citations in response to the legal research question. (And, it cited to 12 cases in the final version of its memo.)

It did a really good job of analysis too!

Generating a Legal Memorandum

Writing a memo was not part of the class assignment because the ChatGPT family was refusing the last few weeks,* and Bing Copilot had to be tricked into writing one as part of a short story, but after seeing Claude 3’s research/analysis results, I decided to just see what happened. I have many elaborate prompts for ChatGPT-4 and my legal memorandum GPTs, but I recalled reading that Claude 3 worked well with zero-shot prompting and didn’t require much explanation to produce good results. So, I decided to keep my prompt simple – “Please generate a draft of a 1500 word memorandum of law about whether Snurpa is likely to prevail in a suit for false imprisonment against Mallatexaspurses. Please put your citations in Bluebook citation format.”

From my experience last week with Claude 3 (and prior experience with Claude 2 which would actually answer questions), I knew the system wouldn’t give me as long an answer as requested. The first attempt yielded a pretty high-quality 735-word draft memo that cited all real cases with the correct citations*** and applied the law to the facts in a well-organized Discussion section. I asked it to expand the memo two more times, and it finally produced a 1,446-word document. Here is part of the Discussion section…

Implications for My Teaching

I’m thrilled about this great leap forward in legal research and writing, and I’m excited to share this information with my legal research students tomorrow in our last meeting of the semester. This is particularly important because I did such a poor job illustrating how these systems could be helpful for legal research when all the compared systems were producing inadequate results.

However, with my administrative law legal research class starting tomorrow, I’m not sure how this will affect my teaching going forward. I had my video presentation ready for tomorrow, but now I have to change it! Moreover, if Claude 3 can suddenly do such a good job analyzing a fact pattern, performing legal research, and applying the law to the facts, how does this affect what I am going to teach them this semester?

*Weirdly, the ChatGPT family, perhaps spurred on by competition from Claude 3, agreed to attempt to generate memos today, which it hasn’t done in weeks…

Note: Claude 2 could at one time produce an okay draft of a legal memo if you uploaded the cases for it, that was months ago (Claude 2 link if it works for premium subscribers and Google Drive link of cut and pasted chat). Requests in January resulted in lectures about ethics which resulted in the above-mentioned cancellation.

Welcome Perplexity AI to the Law Library! Move Over Google

On a couple of podcasts, I’ve heard a lot of hype about Perplexity AI and how it could be a big competitor to Google. Even though I really like new, generative AI things, it still sometimes takes hearing about something multiple times before I overcome inertia and finally check it out.

While attempting to write a blog post about whether the memory-impaired ChatGPT-4 could still perform well on a mock bar exam (spoiler alert – my tests so far indicate it can!), I was Googling for information. Specifically, I was searching for articles around the time GPT-4 Passes the Bar Exam was published on SSRN and some background on the paper author’s methodology. It was taking a long time to piece it all together… Then suddenly, I overcame my laziness and decided to check out Perplexity AI. When I reached the site, I realized that I had actually used it before! For whatever reason, I found it more appealing the second time!

Question: Tell me about how ChatGPT-4 passed a mock bar exam and the methodology that was used to arrive at that conclusion. (Note: Click here to view the full answer on the system.)


Question: Tell me about how ChatGPT-4 passed a mock bar exam and the methodology that was used to arrive at that conclusion.


Watch out Google! I really love access relevant information is getting so much easier.