Leadership, Ethics, and Lifting Weights: Teaching Persuasive Legal Writing in the AI Era DOWNLOAD PDF

In the world of legal writing instruction, we’re in a transition not unlike the one agricultural workers faced when they moved from animal-powered to steam-powered farming equipment. In the pre-industrial era, many humans had agricultural jobs that involved significant physical labor. Some built muscles and motor memory naturally as they guided a horse-drawn plow. Then the industrial and technological revolutions arrived. Nowadays, not all agricultural workers build muscles while earning a paycheck. Instead, the option is available for some to sit in the air-conditioned, GPS-guided cab of a high-tech tractor, and go to the gym after work to build muscle instead. 

Generative artificial intelligence (GAI) has changed and will continue to change legal writing instruction in a similar way. In the past, students who were assigned to draft memos learned memo-writing skills along the way. They naturally had to refer to examples, read cases, and ask questions as they worked to put together each section of a brief. Now, in the age of GAI, memo-writing skills may not be built naturally when completing course assignments. Students may be using GAI (either sanctioned or unsanctioned by the course policies) to complete brainstorming, researching, writing, revising, and polishing tasks. If they so choose, they can refer exclusively to GAI-drafted examples. They may read no cases in full; they may rely on GAI-created extracts or summaries alone. They may not come to the instructor with questions because GAI’s answers are far more timely and seemingly definitive.

This change in the legal writing landscape requires a response similar to the one made by those who spend long hours in a tractor cab each day but still want or need to build muscles. Lifting some free weights while monitoring the tractor’s automatic guidance system is probably not the most efficient approach. Choosing a time and a place to work on strength training makes more sense. Similarly, structuring curriculum to promote the mastery of important lawyering skills is likely the most efficient response to the GAI era. (And it is certainly not a new idea—many legal writing instructors have taken this approach even before the age of GAI!)

And yet, even though it doesn’t naturally build muscles, tractor operation comes with risk. In fact, “[t]ractor overturns are the leading cause of fatalities in the agriculture industry.” To reduce tractor overturn risk, farmers should have a “roll-over protective structure and a seat belt” and avoid driving in a “risky manner.” As we know, students (or practitioners) who start using GAI tools for the first time to research or draft face risks as well. Thankfully, they probably won’t be physically injured, but without the metaphorical equivalent of protective structures, seat belts, and safe driving lessons, things still can go ethically or analytically upside down.

Legal writing instructors can update curriculum for the GAI era with these considerations in mind. There are at least three key frameworks to consider when making updates. The first is legal ethics. Independent professional judgment is one of the most important duties to introduce to students, even during their 1L year. Students also need to develop skills around GAI supervision, as they are likely to need those skills in some capacity during their careers. The second framework is skill building. Students need to be given opportunities to develop writing skills in GAI-resistant or GAI-aware ways, so we are confident they are truly developing the skills they need to be successful going forward. The third is professional identity formation. Professional identity formation involves students internalizing the “characteristics, values, and norms” of the legal profession. In particular, this essay emphasizes leadership development, an important piece of making good decisions about when and how to use GAI. Although the way these frameworks play into any individual course will vary, this essay provides a case study describing how the frameworks were implemented into one spring semester 1L persuasive writing course.

Introducing ethical generative AI use

One example of the ways students built their metaphorical muscles in the course was through an ethical GAI use module and corresponding graded application quiz they took at the beginning of the semester. Here are the components students received:

• A short reading on the model rules of professional conduct and how they apply to GAI use;

• A model law firm GAI use policy; and

• A decision tree for evaluation of a point generated by GAI.

   

The decision tree looked like this:

Step 1: Check Your Qualification

• Am I qualified to evaluate this sentence?

  • If YES: Go to Step 2.

  • If NO: Ask: 

    • Is it efficient to qualify myself to evaluate it?

If YES: 

Create a step in your research plan to qualify yourself (e.g., legal or factual research).

If NO: 

Scrap the sentence.

Step 2: Decide to Keep or Scrap

• Keep the sentence ONLY if you can provide evidence for ALL three items: 

Accuracy: Evidence that the sentence is legally or factually correct. Accurate means: it (1) is true, (2) is followed by a citation to a real authority, (3) is supported by the authority cited, (4) has an appropriate signal, if needed, and (5) has quotation marks around phrases of five or more words that are the same as the original source.

Persuasive Value or Disclosure Requirement: Evidence that the sentence is useful for the client’s argument (persuasive) OR required to disclose.

Scope: Evidence that the sentence falls within your assignment’s scope.

• Scrap the sentence if ANY of the three items above is missing.

Step 3: Repeat

• Move to the next sentence and start back at step 1.

            Students took a short quiz where they applied what they learned from each of these three components (the reading, the law office policy, and the decision tree) to hypothetical scenarios that took place at a law office.

Reinforcing skills with spaced repetition

I had students make decisions around GAI during multiple steps of the simulated litigation process during the spring semester. This spaced repetition—practicing tasks throughout the semester with time gaps in between—can help reinforce both (1) the student’s professional identity as a leader and supervisor, and (2) critical prompting and evaluation skills. The course was designed to start with the complaint. Students then drafted a preliminary injunction brief, drafted a summary judgment brief, completed oral argument, and ended with a transactional assignment. GAI was organically incorporated at many steps in the process. For example, at the beginning of the semester, we uploaded the Complaint to Westlaw’s Litigation Document Analyzer and evaluated the results as a class.

Throughout the semester, I also reminded students that they do not need to draft using GAI. Just because we practice a strategy in class does not mean it’s the strongest approach for any particular piece of legal work. The goal is for students to develop the judgment to decide when to use GAI and when to use human power.

Forming professional identity through a GAI-generated CREAC conference

One significant GAI checkpoint assignment involved students using GAI to draft a persuasive CREAC and then evaluating it in a conference with the instructor. The assignment was designed to help reinforce the professional identity formation framework. I want students to approach GAI output from a place of leadership and skepticism, building a supervisory identity. It is easy for 1L students to feel like a GAI tool knows more than they do, since its output is so instantaneous and confident in tone. It also typically uses correct grammar and is formatted professionally—all things we traditionally associate with substantive credibility.

But as we know, the uniform style and typeface of a GAI-generated output can mask significant errors and massive discrepancies in the quality of the output. A single human tends to produce writing with consistent quality, at least when writing in a single domain area. GAI can produce high quality, low quality, and medium quality writing and information about law in the same output, in different outputs several hours apart, or in outputs provided to two students with the same prompt during the same class. One sentence could contain an accurate statement of law while the next could contain an inaccurate statement of law. Students may not realize this unless we tell them. Working with GAI outputs requires different editing strategies than when we are editing ourselves or other humans. Knowledge of these characteristics of GAI outputs helps students build GAI literacy.

In my class, students needed to send me three things before the writing conference. The first thing students sent was the GAI-generated CREAC. I asked them to prompt GAI to generate a CREAC on an issue they had already briefed in their first memo; this meant they had expertise to evaluate the output. They were supposed to do their best to make it 2-4 pages long, include citations, include all the parts of CREAC we had discussed in class, and be persuasive in tone. We had practiced several prompting strategies in class, and they could use any of them for this task, including attempting to prompt and re-prompt to draft the entire CREAC at once, or to go section by section. They had to prompt at least five times. I asked them to give me just the GAI-generated draft, with no human edits. The second thing students had to include was their written evaluation of three of the sentences from the GAI-generated CREAC (including the citations) using the decision tree above. Finally, the third thing they included was a list of five things to change or cut from the GAI-generated draft. They could be large or small things. Students were given a list of questions to consider when deciding what to change or cut, including:

• Is the draft in CREAC format?

• Is the draft complete? What’s missing?

• Is there anything in the draft that’s outside of the scope of the assignment? 

• Does the draft put your client’s best foot forward?

• Does it incorporate a persuasive theme?

• Does every sentence make an accurate statement of law or fact?

• Does the application section mention every legally significant fact related to the prompted issue?

• Does the tone sound like your voice, or is it flat and mechanical, flowery and academic, or some other tone that doesn’t sound like you?

• Are any direct quotes in quote marks? Do they quote the EXACT words of the underlying document or opinion? Is statutory language quoted?

• Are the paragraphs in a logical order? Does each paragraph have a topic sentence? 

• Do you find the draft persuasive?

• Do you like it?

• Are all the legal citations correct? Are they to real cases? Are they the cases you want to cite? Are there accurate pincites?

• Are the fact citations correct? Are they to the documents you want to cite?

The question was not “is this good enough?” I prepared them for the reality that it wouldn’t be. Instead, the question was: “what do I need to cut or change?” I hoped it would help them see the labor that would be required to evaluate a GAI-generated CREAC. It also helped align our perceptions of how accurate the GAI-generated sentences were. 

This assignment shows how much professional identity formation and GAI instruction have to offer each other. Many students enter law school with leadership skills from previous jobs, academic, extracurricular, or sports experiences. I love having the opportunity to reinforce a skill that students may already feel good about during a year when so many things are new and challenging. Supervising GAI outputs as described in this conference assignment can help students build or reinforce the professional identity formation skill of leadership. And effective GAI supervision requires strong leadership. Pairing practice of these two skills in a single assignment can be a win-win.

Designing a GAI use policy that reinforces all three frameworks

I set policies for both collaboration and GAI use in the course on the assignment level. The GAI policies were set for specific reasons. Generally, when GAI was allowed, it was allowed so students could practice using it ethically and developing professional identity. When it was banned, it was to give students the opportunity to isolate and refine specific cognitive skills, just as one would isolate and exercise a specific muscle during a workout. Students were generally allowed to use GAI tools for any purpose on most assignments. I tried to explain the rationale behind why GAI was permitted on some assignments and prohibited on others. Where appropriate, I also explained which GAI strategies I thought would be successful and which would not. To avoid confusion, I used the same permissions box at the top of each assignment sheet. It explained in words how GAI could be used and it also used color to visually signal whether it was permitted. For each major assignment, students were asked to add a GAI notification describing which tool(s) they used, what the purpose was (brainstorming, researching, drafting, revising, or editing), and acknowledging that the student was responsible for the accuracy of the final product.

One thing I did during this semester was to provide an initial survey to my students about their experiences with and feelings about GAI. I want to be respectful of students who might be conscientious objectors to using GAI and plan to be flexible about my approach to requiring students to use it for this reason. Although many concerns about GAI’s impact on society also give me pause, I ultimately thought it was important to make sure students had practice using it in a lower-stakes environment before they go off to summer jobs. This is valuable for those whose employers have defined GAI use policies and can provide students with direction about when and how to use GAI. It is potentially even more important for employers who are GAI novices themselves and may not have thought through a GAI policy for their law office.

Conclusion

I know I will continue to make changes to my curriculum going forward to help students get practice with these three frameworks. In the future, I may have students complete three writing assignments in the spring semester. At least one would be in a proctored class environment (the “gym”) so I can best assess whether they are able to independently execute the writing skills we are learning. One would be an assignment they write in and out of class with GAI assistance, to practice using GAI and exercising lawyerly judgment in GAI-related decision making (the “independent judgment”). The third would be an editing assignment in which they would have an opportunity to focus on the supervision skills they will need to supervise GAI or other humans in the future (the “leadership”). As we face rapid ongoing technological change, informing curriculum design with a foundation of professional identity formation, ethical awareness, and verified lawyering skills will help students succeed in their summer internships and beyond.

1. ^

    See, e.g., Top tips to keep your tractor air con in order, Farmers Wkly. (June 30, 2012), https://www.fwi.co.uk/machinery/top-tips-to-keep-your-air-con-in-order (“With farmers spending more hours in the cab than ever before, fixing up the air-con should rank highly on the pre-harvest to-do list.”); Zděnek Kvíz & Milan Kroulík, Automatic guidance systems in agricultural machinery as a tool for drivers’ mental strain and workload relief, 63 Rsch. in Agric. Eng'g S66 (2017). This metaphor is intended to be illustrative only. Farming is hard work, with or without a fancy tractor, and I am very thankful for the important work of those in agriculture.

2. ^

    See, e.g., The Derek Bok Center for Teaching and Learning, Designing Courses & Assignments in the Age of AI, https://bokcenter.harvard.edu/courses-and-assignments-in-age-of-ai (last visited Mar. 20, 2026) (explaining that take-home papers and essays “present a high risk of being completed fluently by AI without detection”).

3. ^

    See, e.g., Carolyn V. Williams, Bracing for Impact: Revising Legal Writing Assessments Ahead of the Collision of Generative AI and the NextGen Bar Exam, 28 J. Legal Writing Inst. 1, 2 (2024).

4. ^

    OSHA, Protecting Agricultural Workers from Tractor Hazards (Oct. 2015), https://www.osha.gov/sites/default/files/publications/OSHA3835.pdf.

5. ^

    Id.

6. ^

    See, e.g., Kyle Jahner, AI Blunder in ‘Big Boy Court’ Prompts DOJ Firing of Attorney, Bloomberg Law (Mar. 13, 2026), https://news.bloomberglaw.com/ip-law/doj-fires-lawyer-who-filed-ai-brief-after-poor-court-showing.

7. ^

    See, e.g., ABA Standing Comm. on Ethics & Pro. Resp., Formal Op. 512, at 4 (“[L]awyers may not abdicate their responsibilities by relying solely on a GAI tool to perform tasks that call for the exercise of professional judgment.”).

8. ^

    See, e.g., University of Chicago, Strategies for Designing AI-Resistant Assignments, https://genai.uchicago.edu/resources/faculty-and-instructors/strategies-for-designing-ai-resistant-assignments (last visited Mar. 30, 2026).

9. ^

    See, e.g., MIT Teaching & Learning Lab, AI-Aware Implementation: Assessment Examples, https://tll.mit.edu/teaching-resources/course-design/gen-ai-your-course/ai-aware-implementation-examples/ (last visited Mar. 30, 2026).

10. ^

     What Is a Law Student’s Professional Identity and What Is Professional Identity Formation? – A Short Introduction, Holloran Ctr. 1, 1 (Sept. 2022), https://law.stthomas.edu/_media-library/documents /holloran-center/professional-identity-definition1.pdf (citing Richard L. Cruess et al., Reframing Medical Education to Support Professional Identity Formation, 89 Acad. Med. 1446, 1447 (2014)).

11. ^

     See Marjorie M. Schultz & Sheldon Zedeck, Predicting Lawyer Effectiveness: Broadening the Basis for Law School Admission Decisions, 36 L. & Soc. Inquiry 620, 630 (2011) (listing “[o]rganizing and [m]anaging [o]thers” as a lawyering effectiveness factor).

12. ^

     The decision tree was edited for concision with the assistance of a GAI tool, Microsoft Copilot.

13. ^

     This first step was inspired by Kirsten K. Davis, J.D., Ph.D., Professor of Law, Stetson College of Law who mentioned starting with checking one’s qualifications during a Gen AI Convo Group webinar in the November/December 2025 time frame.

14. ^

     See, e.g., Tracy L. M. Norton, The Supervision Gap: AI And The Emerging Professional Identity Formation Obligation That Begins in Law School’s First Year, 31 J. Legal Writing Inst. ___, 24-25, 45-47 (forthcoming 2027), available at https://dx.doi.org/10.2139/ssrn.6471718.

15. ^

     See, e.g., Brian Sites, Learning Theory and the Law: Spaced Retrieval and the Law School Curriculum, 43 Law & Psychol. Rev. 99, 105 (2019) (“Spaced repetition offers learning benefits both for direct factual recall and for applying information.”).

16. ^

     Lawyerly judgment invokes the legal ethics of maintaining independent professional judgment and the professional identity of the lawyer as leader and supervisor who does not abdicate decision making to GAI. See, e.g., Jessica Lynn Wherry & Frances C. DeLaurentis, Becoming Our Best: Professional Identity Formation as a Pathway to Lawyering Excellence in the Age of Generative AI, 57 St. Mary’s L.J. 791, 820, 827-31 (2026) (discussing independent professional judgment as an ethical obligation, a skill, and a professional identity formation “virtue”).

17. ^

     This idea was introduced to me by Dr. Ryan Watkins. See Ryan Watkins, From AI to A+: Prepare Your Students for Using ChatGPT and other AI, Medium (Jul. 27, 2023), https://medium.com/@rwatkins_7167/from-ai-to-a-prepare-your-students-for-using-chatgpt-and-other-ai-6ecbfb14297b.

18. ^

     I was delighted to find that this approach is consistent with the one taken by the Eighth Edition of the ALWD Guide. See Carolyn Williams, ALWD Guide to Legal Citation 469 (8th ed. 2026).