Back to Publications

The Second Draft -

Using a Comparative AI Exercise to Improve Student Drafting DOWNLOAD PDF

  • Karen Mika
    Senior Legal Writing Professor
    Cleveland State University College of Law

Published: December 2025

It is hard to dispute that the integration of Artificial Intelligence (AI) is actively reshaping both legal education and the legal profession.[1]  We are currently in a dynamic developmental phase, not unlike the development of computerized legal research.[2] Currently, students and attorneys have been presented with a multitude of AI tools that do slightly different things in different ways.  The rapid integration of AI presents a serious challenge for the legal field. Although the tool's upsides are too significant to be ignored, its known downsides, such as 'hallucinations,' the 'garbage in, garbage out' nature of prompting, and the risk of it becoming a crutch, mean it cannot be responsibly adopted as a new standard without careful consideration.[3]

Almost everyone who has worked with AI in the legal profession agrees it has a significant future. However, they also acknowledge that the technology is currently deficient in key areas of legal practice and education.[4]  Experience has demonstrated that AI offers distinct advantages for specific facets of legal work, although it is not universally beneficial for all tasks.[5] AI has proven especially useful for aspects of editing and more rote tasks that do not demand deep critical thinking or judgments about prioritizing the “best” information to convey to a reader. This remains true even as AI continues to improve in that area. Nonetheless, most legal educators have already observed AI's capacity to organize and suggest improvements in the clarity of writing in any type of legal document drafting, whether litigation-oriented or transactional.[6]

One of the places where AI use can be adopted beneficially in the classroom is in drafting courses.[7] Rather than viewing AI solely as a potential tool for academic dishonesty, in drafting courses, legal educators can harness it to enrich learning and practical applications, while eliminating some of the risk that would come from experimenting with AI in a law firm environment. This article describes an exercise I currently assign in my upper-level General Drafting course.  The course, which requires drafting clauses in various subject areas[8], integrates AI in each exercise.  It asks students to first draft without the help of AI, use AI to review their original drafts, and then critique both the AI product and their own work.  The following is an example of one of the exercises.

1. The Exercise

1.1 Step 1: The Drafting Prompt

In my course, I structure drafting exercises to teach students how to organize material to suit the field of law in which they are writing.  Each exercise begins with a unit of instruction that I provide.  It includes background law, organizational materials, and samples the students may rely on for getting a general idea of organization and expectation.  They are also provided basic instruction in drafting prompts in AI.

Students must then draft clauses in the assigned field based on the materials they are given.  It is after this that students are asked to use AI[9] to re-draft their clauses and reflect on what the outcome is.  I ask students to reflect on the output and whether it was better, the same, or worse than their clause.  I also ask them to address what aspects of the prompt may have caused the AI output and how the prompt might be improved.  Finally, I provide feedback on their AI drafting experience. This includes suggestions for improving their prompts and observations on how the AI output itself could have inspired further additions or clarifications to the rewritten clause.

For one exercise, the instruction below was provided as part of an exercise in drafting clauses from a typical Landlord/Tenant Agreement:

Please draft the following:

A parking clause for a tenant lot that has 40 spaces. Each tenant has one assigned space and can pay an extra $25 a month for a second assigned space, but there are only 5 secondary spaces available. After that, the tenant must go on a waiting list and park in one of the free spaces. Assume there are 20 tenants in the complex. Assume that guests may use the lot, but not the assigned spaces and guests may not use the lot overnight for more than 5 days without a penalty. As you design the clause, consider the following:

 

- Any permissions you might want for guest parking.

- Any penalty system for violating the parking lot rules, including a reporting system and recurring violations.

- Any official documentation that must be submitted to a central office.

- Anything else you think necessary to consider.

1.2 Step 2: The Initial Student Draft

The first step is for the student to create an original draft from scratch. Students should apply their knowledge from the materials, which include instruction in foundational law, organization, and tips for clarity. Those instructional materials should allow the student to make their own initial informed decisions about structure and substance, creating a baseline of their own work. 

The following is a representative example of a student's draft[10], completed without AI assistance:

Parking Clause

General Distribution of Parking Spaces The parking lot has 40 total spaces, 20 of which will be assigned to tenants as primary parking. There are 5 spaces available as secondary parking. Tenants may rent one of these spaces for $25 per month on a first-come, first-served basis. If all secondary spots are reserved, tenants may join a wait list. There are 15 guest parking spots available on a first-come, first-served basis. These spots are available for use by guests of any Tenant.

Registration and Application Application for primary parking and secondary parking must include vehicle make/model, license plate, and apartment number. Registration for guest parking must include vehicle make/model, license plates, apartment number being visited, dates of stay, and phone number of driver.

Guest Parking Guests may park only in guest parking spaces. Overnight guests must register their vehicles with the management office. Registration requires license plate number and duration of stay. Guests may not stay more than 5 days per month.

Reporting and Violations A Tenant who believes a violation has occurred may report their suspicion to the management office which shall investigate and take any necessary action. Penalties for parking violations include (at discretion of management), but are not limited to:

* Verbal or written warnings 

* Ineligibility for secondary parking or guest parking.

* Vehicles may be towed at vehicle owner’s expense.

1.3 Step 3: The AI-Generated Draft for Comparison

Here is an example of the AI output using only the assigned exercise as the prompt:

PARKING

The [Apartment Complex Name] parking lot consists of 40 total parking spaces, located on-site and designated for use by residents and their guests under the following terms and conditions:

1. Assigned Parking * Each unit is assigned one (1) reserved parking space at no additional cost.

* Tenants may apply for a second reserved parking space at an additional cost of $25.00 per month. Only five (5) secondary spaces are available at any given time.

* Secondary spaces will be assigned on a first-come, first-served basis, and tenants must complete and submit a Parking Addendum Form to the Leasing Office for approval and assignment.

* Once all five secondary spaces are assigned, tenants may request placement on a waiting list. During the waiting period, tenants may only use unmarked guest spaces.

2. Guest Parking * Guest vehicles are permitted to use any unmarked and unassigned parking spaces.

* Guests may not park in any assigned tenant space, regardless of vacancy.

* Guest vehicles may park overnight for up to five (5) consecutive days within any 30-day period. Any guest vehicle found parked overnight beyond this time without written authorization from the Landlord is subject to a $50.00 daily penalty and may be towed at the guest’s or Tenant’s expense.

* Tenants are responsible for informing their guests of all parking rules.

3. Violations and Enforcement * Unauthorized use of an assigned space by a guest or another tenant may result in: A $50.00 fine per incident for the offending tenant, and Towing of the unauthorized vehicle at the owner’s expense.

* Tenants who receive three (3) or more parking violations in a rolling 12-month period may: Lose eligibility for a secondary space (if applicable), or be subject to lease penalties, including potential non-renewal.

* All violations must be reported to the Leasing Office in writing, via email or a completed Parking Violation Report Form, available in the office and on the tenant portal.

4. Documentation and Registration * Tenants must register all vehicles (including guest vehicles staying more than 24 hours) with the Leasing Office. Registration must include: Make, model, and license plate number, and proof of current registration and insurance.

* Each registered vehicle will receive a Parking Permit Tag. Vehicles without tags may be deemed unauthorized.

5. Prohibited Conduct * No vehicle may block entrances, dumpsters, fire lanes, or other vehicles.

* No vehicle maintenance, washing, or oil changes may be performed on-site.

* No trailers, RVs, boats, or oversized vehicles may be parked in the lot.

1.4 Step 4: The Comparative Analysis: Improving the Original Work

This next stage is the core of the learning exercise. By placing the two drafts side-by-side, the student can identify concrete opportunities to improve their original work. The following are comparative assessments that a student could have made demonstrating how, in this situation, the AI draft includes some aspects superior to the original student draft.

Structure and Clarity

The student's draft is functional, but its organization could be improved. For instance, registration requirements are split between two sections, "Registration and Application" and "Guest Parking."  Although the student’s draft generally covers the required categories, there is overlap between them. The AI version separates out many more categories, which makes the entirety of the clause clearer.   Moreover, in the non-AI version, the placement of the penalties section after the Guest Parking section makes it unclear whether the penalties relate only to the Guest Parking rules or the permanent space rules.  Although the AI version does not change the placement of the penalties section, it makes it clear that the penalties apply to both a violation of the Guest Parking rules as well as the permanent space rules.

Specificity and Detail

A key weakness in the original student draft is the lack of specificity. The "Reporting and Violations" section is vague, stating that management "shall investigate" and that penalties are "at discretion of management." This creates ambiguity. The AI draft, however, provides specific, enforceable terms:

  • It establishes a clear penalty of a $50.00 fine per incident.

  • It creates a "three-strikes" rule for repeat offenders, which could lead to non-renewal.

  • It formalizes the reporting process with a "Parking Violation Report Form."

This comparison can potentially teach the student an example of how to move from general statements to concrete details required for an effective contract or policy.[11]

Identifying Omissions 

The student's draft correctly addresses all the items in the prompt. However, it fails to anticipate other potential problems. As an example, the AI draft demonstrates better anticipation of foreseeable problems by including a "Prohibited Conduct" section. This section thoughtfully adds rules that prevent tenants from performing oil changes on-site, parking large RVs or boats, or blocking fire lanes. Seeing this, the student could learn a valuable lesson: a good drafter doesn't just answer the client's direct questions but also anticipates future issues to create a more comprehensive and helpful document.

Overall

It should be noted that, for some students, the AI exercise did not necessarily write a superior clause.  In some instances, the AI product was merely “different” in terms of organization, emphasis or foreseeable issues.  However, my goal in the exercise was that the student would use AI to notice where they could improve the work and what other ideas they could incorporate into the original clause. Furthermore, I hoped the students might conclude that further “conversation” with AI about the clause would lead to other improvements.[12]

1.5 Step 5: Self-Reflection

An equally important part of the AI exercise is the self-reflection component. After completing the AI usage part of the exercise, the students were asked to reflect on their experience with AI. They were asked to assess their own prompts and to evaluate whether AI did a better job in accomplishing the assigned task.

In the example above, many students reached the conclusion that the AI output was better than the original composition and that they would not have recognized missing features or helpful expansions without seeing the AI output. Many students also remarked that it would have been beneficial to continue having a discussion with AI rather than just putting in one or two prompts and receiving an answer.  Most students also concluded that although AI was a beneficial tool, they did not believe it replaced the need for a human being who would be able to verify information and correct errors (an issue addressed in more detail below).

2. Necessary Cautions for Using AI in Drafting

Although the exercise I use demonstrates AI's utility, it is also crucial that the professor emphasize at the outset that AI should be used as a tool to assist, but not to replace, legal judgment or confirmation of accuracy.  Most students recognize this after reflecting on their AI revised clauses; however, many students do believe AI will provide a quality answer for any legal question asked.  The professor must consistently remind students that all AI output must be reviewed for accuracy and scrutinized to ensure the output meets all expectations. 

For example, the AI draft above created ambiguity in the guest parking rules by requiring guest registration for vehicles staying more than 24 hours but not providing when registration should occur. This discrepancy could potentially be ambiguous when calculating when the guest vehicle violates the rule for being parked in the lot more than five nights.

There are multiple other aspects of the AI draft that could have proved problematic:

The AI model suggested a $50 fine, but it has no knowledge of local ordinances or whether there are prohibitions for penalties in a given jurisdiction.[13] The student must understand that their job is to research and confirm that such a provision is legally enforceable.

The AI model has firm parking lot penalties, including the potential for lease penalties; however, these penalties (especially any eviction) may actually be a breach of contract in relation to the full lease, particularly if that lease includes notice requirements. The student would need to evaluate these potential inconsistencies.

The AI model requires that vehicles in the lot have insurance. However, different states have different requirements for insurance. The student would need to research to ensure what the financial responsibility requirements of the state are. The agreement would also need to include a provision about how proof of insurance would be provided and the timeframe for periodic updates. 

3. Conclusion

Although AI is a beneficial drafting tool, ultimately, the lawyer or student is fully responsible for the final work product. AI can be used to generate ideas and structures, but the human drafter must perform the final analysis, ensure accuracy of both content and law, and tailor the document to the client's precise needs.


[1] Alexandria Serra, AI Lawyering Skills Trainers: Transforming Legal Education with Generative AI, 16 Case W. Res. J.L. Tech. & Internet 74, 74 (2025); see also Tatia Dolidze, The Evolving Role of Artificial Intelligence in Legal Education and Research, 11 Law and World 92, 92-93 (2025).

[2] See generally Carolyn Williams, Bracing for Impact: Revising Legal Writing Assessments Ahead of the Collision of Generative AI and the NextGen, 28 J. L.W.I. 1 (2024).

[3] See Dolidze, supra note 1, at 93,94; see also Fedbarblog, The Times They Are A-Changin’: The Rise of Generative AI in the Legal Profession, Federal Bar Association, May 7, 2024, https://www.fedbar.org/blog/the-times-they-are-a-changin-the-rise-of-generative-ai-in-the-legal-profession/; Joseph Regalia, From Briefs to Bytes: How Generative AI is Transforming Legal Writing and Practice, 59 Tulsa L. Rev. 193, 193 (2024).

[4] Marjorie Richter, J.D., How AI is Transforming the Legal Profession, Thomson-Reuters, Aug. 18, 2025, https://legal.thomsonreuters.com/blog/how-ai-is-transforming-the-legal-profession/.

[5] John Villasenor, Generative Artificial Intelligence and the Practice of Law: Impact, Opportunities, and Risks, 25 Minn. J. L. Sci. & Tech. 25, 25-26 (2024).

[6] Id.

[7] Drafting course, often taught in the upper level at law schools, offer students more of an opportunity to improve prompts, organization, and the specificity of language, often without having to involve extensive case research.

[8] The subject areas covered are: simple wills, landlord-tenant agreements, employment agreements, general releases, demand letters, and informational client letters. The course also includes lessons in grammar and the clarity of language.

[9] Students may select whatever AI program they would like to use. No specific AI program is assigned. 

[10] The sample is an amalgamation of various clauses drafted by students over several years. It is not the work product of one student.

[11] To be sure, increased specificity is not always desirable. For example, I teach my students that landlords might in some instances jockey for less contractual specificity in order to maximize their discretion about how to enforce certain penalties.

 

[12] Although my students are currently novices with respect to using AI, they were given examples of how using AI as a conversational tool could help improve the output. Some students replicated this usage, while other students simply took the “one and done” route.

 

[13] Many jurisdictions and localities have statutes that cover both fees and penalties for a variety of landlord-tenant items, including raising rent, registration fees, and late rent. The trend in the more tenant-friendly jurisdictions is to prevent certain “unreasonable” charges. For example, the City of Bellingham, in Washington state, has local ordinances prohibiting excessive fees and deposits in residential rentals. See Bellingham, Wash., Mun. Code §§ 6.16.010–.120 (2025).