ChatGPT for Lawyers and Legal Professionals — 102 (Prompts)

September 17, 2023

Blog Img

In this article we delve into the bedrock of a successful ChatGPT query — ‘prompt engineering.’ Uncover a plethora of tools and methods to design and execute effective prompts, elicit accurate responses and beyond. Get ready for a transformative journey.

In the continuation of my prior post, where I left you with an eager anticipation, I’m thrilled to fulfill my commitment by presenting a hands-on tutorial that delves into the art and science of crafting effective prompts for ChatGPT and analogous LLMs like Bard.

So, without much ado, let’s get started.

How do you write the best Chatgpt prompts

As you can see in the image above, I gave Chatgpt 3.5 a simple ‘prompt’ ‘How do you write the best Chatgpt prompts’.

What it did was to visit its vast backend knowledge repository and locate the most relevant response based on the prompt query and suggest a concise relevant answer.

It is to be undestood that since it is trained on the repository that might be contain incorrrect information too, the response though invariably correct is not necessarily entirely accurate.

But it is always good to get a first draft in seconds as opposed to devoting a considerable length of time and effort to just get started.

That’s precisely the approach we’ll adopt with ‘prompts’ to initiate our legal research process.

Prompts for Legal Research A simple starting prompt could be,

‘What is the relevant legal topic or issue?’

Additional context can be provided, such as requesting, ‘Please provide a list of case laws, statutes, and regulations related to this topic.’

For more detailed exploration, you might inquire about a specific statute or case law, like ‘Please provide an in-depth analysis of the ‘Balfour vs Balfour’ case law.’

In response, a comprehensive explanation can be provided, encompassing a brief background, a summary of the case’s facts, key issues, the court’s ruling, and the legal principle established.

Now if you would like to need the same information very briefly, you could ask for it to be brief as in ‘within 10 sentences’ or 300 words or a measure of length.

I just asked Chatgpt to be brief so it provided me with this response:

In a similar vein, a multitude of prompts can facilitate rapid legal research and analysis.

I invite you to stay tuned for my upcoming article, where I will present a categorized compilation of effective prompts tailored to various legal tasks.

For now, let’s move to the next important legal task.

Prompts for finding Precedents

Finding precedents of established legal principles, rules, or tenets in law, whether as case laws or customs or usage, is an important task for any legal professionals.

The reason being that such legal precedents help establish consistency, predictability, and justice in judicial decision-making. However, professionals often struggle to accurately and timely locate the relevant precedents.

This is where Legal Language Models (LLMs) prove to be of immense assistance.

Let me give an example of a starting prompt. Also give it a length constraint. Also, let’s introduce a role play of a tax consultant.

Role — Top London Tax Consultant;

Task — Overview of legal precedence in the late filing of HMRC tax returns due to illness (task is specific to UK jurisdiction);

Length of answer — 2 sentence

In a matter of seconds, Chatgpt provides you with a general overview of the legal precedence in this case.

You can ask it to elaborate in ‘bullet points’ and it gives a response as below:

The next prompts can be more specific like ‘ Give me a jist of key cases that exist in this area of law’ or a followup question ‘ Provide the key legal arguments in (case law)’.

I experimented with a prompt to bring out the consequences of misinterpreting the legal precedents. And the response was very precise.

Prompts for drafting Legal Documents

As a legal professional, you are likely familiar with the various legal document drafting software tools available to assist you.

However, the advent of Language Models like GPT-3 has indeed simplified this process further, reducing the necessity for specialized software.

Let us look at some prompts to aid the document drafting process.

‘Write a demand letter between Bob and Mary for compensation and penalty for slipping on garbage including banana peels thrown from the balcony on the road down below in Memphis, Tennessee jurisdiction.’

or better still

‘Compose a demand letter on behalf of Bob to Mary, seeking compensation and penalty for injuries sustained due to slipping on garbage, including banana peels, discarded from the balcony onto the road below, within the jurisdiction of Memphis, Tennessee.’

The reason I have introduced two (2) prompts seeking similar response is to suggest that if the essence of the prompt is generally correct, the actual wordings do not matter as far as it is logically and grammatically coherent.

Now let us see the response from my first prompt.

[Your Name] [Your Address] [City, State, Zip Code] [Email Address] [Phone Number] [Date]

[Recipient’s Name] [Recipient’s Address] [City, State, Zip Code]

Subject: Demand for Compensation and Penalty due to Slip and Fall Incident

Dear [Recipient’s Name],

I hope this letter finds you well. I am writing to address a concerning incident that occurred on [Date] involving my client, Mr. Bob Johnson, and your property located at [Address]. The incident has prompted my client to sustain injuries, incurring medical expenses and emotional distress.

On the aforementioned date, Mr. Johnson was walking on the road below your balcony when he slipped on garbage, including banana peels, that had been negligently thrown from your balcony. The lack of proper disposal of waste resulted in Mr. Johnson’s slip and fall, causing him physical harm and subsequent financial and emotional strain.

Under Tennessee law, property owners have a duty to maintain a safe environment for visitors and passersby. Negligent actions that contribute to hazardous conditions, such as throwing waste onto public areas, can result in liability for injuries caused as a direct result of such negligence. In light of this, I am writing to request the following:

Compensation: We demand that you compensate Mr. Bob Johnson for his medical expenses, loss of income, and pain and suffering resulting from the slip and fall incident. An itemized list of expenses incurred is attached herewith for your reference.

Penalty: Given the severity of the incident and the violation of public safety norms, we also request that you pay an additional penalty as allowed under Tennessee law to deter such actions in the future.

We understand that addressing this matter amicably is in the best interest of all parties involved. We kindly ask that you respond within [reasonable time frame, e.g., 14 days] from the receipt of this letter to discuss the resolution of this issue. Failure to respond or reach an agreement may compel us to pursue legal remedies available to our client under Tennessee law.

Please consider this letter as a sincere effort to resolve the matter without resorting to legal action. We hope to avoid unnecessary legal expenses and inconveniences for both parties. Kindly respond to this letter at [your email address] or [your phone number] to initiate a constructive discussion.

Thank you for your attention to this matter. We look forward to your prompt response.


[Your Name] [Your Title] [Your Law Firm Name] [Bar Association License Number] [Attachments: Itemized list of expenses incurred]

And the response from my second, better worded, prompt.

Chatgpt suggests that there is no change in answer as the essence of the situation and the legal context remain consistent despite slight variations in the wording of the prompt.

I added a further prompt to know if a particular law holds in this case.

This is just one example. One can do similar drafting for contracts, wills, power of attorney forms and even Non Disclosure Agreement (NDA).

Not only that you can go a step further to draft regular legal forms and waiver documents. Let us take the example of a liability waiver.

Draft a liability waiver for Personal Injury with the following terms and condition

Releasor: Bob

Releasee: Mary

Effective Date: Immediate Effect

Event: Personal Injury caused by releasee’s dog running into the releasor’s yard and biting them

Consideration: One time Payment of $10,000

And with a click it drafts a ‘Release for Personal Injury Waiver document’. Of course it could have been more precise if more context and details were also provided such as jusrisdiction.


This Liability Waiver (“Waiver”) is entered into on this [Effective Date], by and between:

Releasor: Name: Bob

Releasee: Name: Mary

Event: The event giving rise to this Waiver is the occurrence of a personal injury sustained by Releasor, Bob, as a result of Releasee’s dog running into Releasor’s yard and biting them.

Consideration: In consideration of the mutual covenants contained herein and for the payment of a one-time sum of $10,000, the sufficiency of which is hereby acknowledged, Releasor agrees as follows:

Waiver and Release of Liability: Releasor hereby waives, releases, and discharges Releasee, Mary, and her agents, employees, successors, and assigns, from any and all claims, liabilities, actions, causes of action, suits, demands, rights, damages, costs, expenses, and compensation whatsoever, whether known or unknown, anticipated or unanticipated, arising out of or related to the aforementioned Event.

Assumption of Risk: Releasor acknowledges and understands the inherent risks associated with having animals, including the risk of personal injury or property damage. Releasor voluntarily assumes all risks associated with the presence of Releasee’s dog in the vicinity of their property.

Indemnification: Releasor agrees to indemnify, defend, and hold harmless Releasee from and against any and all claims, liabilities, actions, causes of action, suits, demands, rights, damages, costs, expenses, and compensation, including legal fees, arising from or related to any claims brought by third parties resulting from the Event.

Entire Agreement: This Waiver constitutes the entire agreement between the parties and supersedes all prior discussions, negotiations, understandings, or agreements, whether written or oral.

Governing Law and Jurisdiction: This Waiver shall be governed by and construed in accordance with the laws of the state of [State], and any legal proceedings related to this Waiver shall be exclusively brought in the courts of [County], [State].

By signing below, Releasor acknowledges that they have read and understood this Waiver and voluntarily agree to its terms.

Releasor: Bob _Signature: ____________________________ _Date: ________________________________

Releasee: Mary _Signature: ____________________________ _Date: ________________________________

Witness: _Name: ________________________________ _Signature: ____________________________ _Date: ________________________________

Note: This is a sample liability waiver for illustrative purposes only. Consultation with legal professionals is advised when drafting legally binding documents.

Prompts for Reviewing legal documents and contracts

Another related legal task that can be efficiently and effectively handled by these models is reviewing legal documents and providing insights into key terms and conditions.

This approach saves time and resources by sparing legal professionals the effort of deciphering complex legal jargon and verbose contracts.

‘Please review of the said contract. Provide a summary of its purpose. Also give a list of three most important terms along with the duration of the contract and the consideration mentioned therein.’

At the end of the prompt we just need to insert the said contract for review. For the purpose of this article, I made up a ‘Confidentiality Agreement’ between Abacus from Edinburgh and Exodus from Glasgow to protect a patent XX to be reviewed by Chatgpt using the above mentioned prompt. You can judge the soundness of the response for yourself.

Prompts to define statutes and regulations for a particular legal issue

It needs to be pointed out that since Chatgpt is trained only on data available till 2021, another LLM Bard might be a better tool for defining statutes and regulations.

To illustrate the example I am using the following prompt — ‘What are the governing statutes and regulation relating to Copyrights in the jusrisdiction of South Korea’

The governing statutes and regulations relating to copyrights in the jurisdiction of South Korea are:

The Copyright Act of 1957 (as amended)

The Enforcement Decree of the Copyright Act

The Fair Use Guidelines for Copyrighted Works

The Copyright Act of 1957 is the primary law governing copyright in South Korea. It protects original literary, artistic, musical, and other creative works, including computer programs, databases, and sound recordings. The Act grants copyright owners the exclusive rights to reproduce, distribute, perform, display, and create derivative works of their protected works.

The Enforcement Decree of the Copyright Act provides more detailed rules on how the Copyright Act is to be implemented. It also sets out the penalties for copyright infringement.

The Fair Use Guidelines for Copyrighted Works are non-binding guidelines issued by the Korean Intellectual Property Office (KIPO). They provide guidance on how the fair use exception to copyright infringement can be applied in South Korea.

In addition to these statutes and regulations, there are also a number of international treaties that South Korea is party to, such as the Berne Convention and the TRIPS Agreement. These treaties provide additional protection for copyrights in South Korea.

Here are some of the key provisions of the South Korean copyright law:

Copyright protection is automatic and does not require registration.

Copyright protection lasts for the life of the author plus 70 years after the author’s death.

Copyright owners have the exclusive rights to reproduce, distribute, perform, display, and create derivative works of their protected works.

There is a fair use exception to copyright infringement.

Copyright infringement is punishable by civil and criminal penalties.

If you are planning to use copyrighted material in South Korea, it is important to be aware of the copyright law and to obtain permission from the copyright owner if necessary. You can also contact the KIPO for more information.

The information provided by Bard is not to be copied verbatim, but rather is a summary of the information that can be found in a variety of sources. It is intended to be used as a starting point for research, and should not be used as a substitute for your own professional wisdom.

Prompts to build a case and develop a legal strategy

Another area where LLMs can be of immense help is in formulating and developing a case. For example, we can use the following prompt for it.

‘Please provide me with a list of all the possible outcomes of a copyright infringement dispute between a South Korean music producer and a foreign streaming platform. Please tell me how the parties can prepare for these outcomes’.

Response using Bard

Response using Bard

Suppose after due deliberation we identify a particular legal strategy for a case. However, we need to make sure of the potential risks and benefits of pursuing such a strategy. LLMs can be used to help you out with this too.

In the above case, say we opt for a ‘negotiated settlement strategy’ and would like to understand the potential benefits and risks in detail. We can use a prompt ‘ What are the potential risks and benefits of pursuing this legal strategy (negotiated settlement). Give me a brief overview for now’

Of course one can keep expanding on this query by giving additional prompts to drill down to the bottom.

Prompt to question witness or conduct discovery responses

‘I am questioning a witness who was seen with my client on the day there was an attempted robbery at my client’s office. Generate a list of questions that could be pertinent to the case. The questions should be concise, straight to the point and clear to elicit a positive response’.

‘Generate a list of questions for opposing parties in discovery for [Employment Discrimination — Alleged Gender Bias in Promotion Decisions]. Please ensure the questions are clear and concise.’

Please describe the promotion process for the position that you were denied.

Were there any specific criteria that were used to evaluate candidates for the promotion?

Were these criteria applied equally to all candidates, regardless of gender?

Can you provide me with the names of the people who made the decision to promote the other candidate?

What were their reasons for making this decision?

Were there any other candidates who were considered for the promotion?

If so, why were they not selected?

Do you have any records of the promotion process, such as meeting minutes or email correspondence?

Can you provide me with a copy of these records?

Are there any other factors that you believe may have contributed to the decision not to promote you?

These are just a few examples, and the specific questions that you ask will depend on the specific facts of your case. It is important to be clear and concise in your questions, and to avoid leading the witness. You should also be prepared to follow up on any answers that the witness gives.

Here are some additional tips for drafting discovery questions:

Be specific. The more specific your questions are, the more likely you are to get the information that you need.

Avoid jargon. Use plain language that the opposing party can understand.

Be relevant. Only ask questions that are relevant to the case.

Be timely. Ask your questions as soon as possible after the event in question.

Be persistent. If you do not get the answers that you need, be prepared to follow up.

By following these tips, you can increase the chances of getting the information that you need to prove your case of employment discrimination.

You can see that Bard has not only provided a list of discovery questions but also given some guidelines as to how to draft good discovery questions.

Prompts to answer client queries

Transform any client queries into prompts for your preferred LLM, leveraging its initial response to provide timely information that can be further enhanced to ensure effective client communication, fostering smooth interactions.


Before concluding this article, it’s crucial to formulate a toolkit for constructing effective prompts. The more specific and precise your prompt, the more enhanced and accurate the response from the AI.

Toolkit for a Good Legal Prompt — a 5 basket approach

As you may have noticed from the prompt examples provided earlier, there are some common traits in each of the prompts.

While the main body changes, the embellishments designed to elicit the appropriate response can be categorized into five main categories.

To keep things simple, let’s refer to these categories as “baskets.”


It is essential that all prompts commence with a strong prime element, such as the context of the query. Give as much information as possible and ask for your desired outcome or result.

For example, “Step into the role of my adversary’s advocate and present rebuttals against the points raised in [legal case].”

In this instance, the prime element “Step into the role of my adversary’s advocate and present rebuttals against the points raised in [legal case]” establishes the context and outlines the desired outcome by asking for counter arguments related to the specified legal case.


It is necessary that the prompt is given a personality, a role, an expertise or a background.

For example, “Embrace the mantle of a seasoned constitutional lawyer examining [legal principle], and expound on how it has evolved over time, shaping modern interpretations.”

In this instance the prompt is being given an identity and legal acumen in ‘a seasoned constitutional lawyer’ that help elicit a more trustworthy and expert response than without it.

PROMPT or steps to be performed

While the term ‘Prompt’ is often used to encompass the entire query, it is actually the essential tasks or steps within a query that orchestrate the formulation of a response.

These core tasks can be regarded as the true ‘prompt’ in practice.

Consider the following _Prompt _as a example.

‘Imagine you are a senior intellectual property attorney specializing in patent law. Given the context of a dispute between Company X and Company Y over a patent infringement claim, provide a comprehensive legal analysis of Company X’s potential liability. Consider factors such as the scope of the patent, prior art, and any potential defenses Company X might raise. Your in-depth legal expertise will guide the formulation of a well-structured argument that addresses the core aspects of Company X’s liability.’

Here, we can break down the prompt into

_Prime _— Given the context of a dispute between Company X and Company Y

_Persona _— Imagine you are a senior intellectual property attorney specializing in patent law.

True Prompt — Consider factors such as the scope of the patent, prior art, and any potential defenses Company X might raise. Your in-depth legal expertise will guide the formulation of a well-structured argument that addresses the core aspects of Company X’s liability

PRODUCT or output or response that you need to elicit

The prompt must be clear about the anticipated output or response, otherwise the generated response could be inadequate.

In the above example, the

product is ‘provide a comprehensive legal analysis’.

The prompt could have been improved for product by specifying the length of the product or response or output. The keywords that can be used here are ‘brief’, 100 words (though it is not very accurate in getting numbers right), in 2 sentences or anything similar.

One can even ask for sources or citations to be included (again LLMs are not very precise as of now).

The key to an excellent product in the prompt is to as specific and precise as possible about the output response that you seek.


Polishing is not precisely a component of the initial prompt but rather the final step in crafting a well-structured prompt.

Following the initial prompt, a sequence of subsequent prompts often follows, which refines, improves, and elaborates on the original query.

These follow-up prompts are based on received responses and are designed to enhance precision, clarify where necessary, and ensure correctness at each stage.

Ultimately, a comprehensive evaluation concludes this iterative process, potentially resulting in a sequence of prompts or signaling its completion

In the upcoming article, I will delve into more instances of legal tasks that can be facilitated using LLMs, exploring advanced prompts.

And now is the moment for a robust disclaimer.

Always bear in mind that the instances provided in this article are primarily hypothetical and illustrative. They may not be applicable in all scenarios. Prior to incorporating any responses into legal work, it’s essential to validate and assess them, applying professional expertise and discretion.