MMOOZZGG.com

AI for lawyers

AI prompts for lawyers: 25 prompts and how to check the answer

6 min read3 sources

AI prompts for lawyers work when the request has a role, a task, source data, an answer format, and a checking rule. According to a Stanford RegLab and HAI study (May 2024), even legal AI research tools make errors in more than 17% of answers, so every prompt below assumes a lawyer checks the result. Instead of client details, the prompts use labels in square brackets: anonymize the text before you send it.

On this page · 4 sections

How a good prompt is built

A prompt, or a request to AI, is a task for an assistant. The more precise the task, the fewer edits you make and the easier it is to check the answer.

1. Role: “You are a corporate lawyer reviewing a contract for the buyer”.

2. The task in one verb: check, draft, compare, find, summarize.

3. Source data: the document text without client details, the governing law, and the party’s position.

4. Answer format: a table, a list of risks, a draft letter, a length.

5. Limits: “Do not invent laws or cases. If you are not sure, say so”.

6. References: ask for the article number, the details of the act, and a quote, so the answer can be checked against the primary source.

25 prompts by task

Replace the labels in square brackets with your own data, leaving out names, details, and amounts that could identify the client.

1. “Review the [type of contract] for the [buyer] under Russian law. Find the terms on liability, deadlines, termination, and jurisdiction, rate the risk, and suggest wording. Text: [text without details]”.

2. “Compare the contract [text] with our standard [standard text]. Table: clause, deviation, risk, what to propose to the other party”.

3. “Draft a disagreement protocol for clauses [clause numbers] of the [type of contract]: our wording and a short reason for each”.

4. “Summarize the contract [text] for the head of [Client] on one page: subject, price, deadlines, liability, exit procedure”.

5. “Find all deadlines in the contract [text] and put them in a table: event, deadline, when it starts, consequence of missing it”.

6. “Build a timeline from the case materials [text]: date, event, document, page. Add nothing of your own, and mark unclear dates”.

7. “Write a 10-point summary of the [hearing transcript], with the page for each point”.

8. “In the parties’ correspondence [text], highlight admissions, promises, and disputes about facts, with a quote for each”.

9. “Draft a claim letter from [Creditor] to [Debtor] under the [type of contract] dated [date]: debt [amount], penalty under clause [number], reply deadline [deadline]”.

10. “Draft a statement of claim about [subject of the dispute] to [name of court] on the facts [facts]. Leave space for the claim amount calculation and references that a lawyer will check”.

11. “Prepare a response to the claim [claim text] for [Defendant]: objections to each demand and the evidence to request”.

12. “List the questions the court may ask our side in the case [short summary of facts] and prepare short answers”.

13. “Draft an outline of a [number]-minute speech on the position [position]: thesis, evidence, reply to the other side’s main argument”.

14. “Name the laws that govern [issue], with article numbers. If you are not sure of a number or version, say so”.

15. “Write search queries for a legal research database on the issue [issue]: keywords, synonyms, period”.

16. “Compare the court decisions [text of the first decision] and [text of the second decision]: where the conclusions match, where they differ, and which facts decided the outcome”.

17. “Translate the contract [text] into [target language] in legal style. Keep terms without an exact equivalent in brackets in the original language”.

18. “Build a bilingual glossary of the terms of the contract [text]: term, translation, comment”.

19. “Write [Client] a letter about the case status [short status]: what is done, what comes next, what we need from them. No legal jargon”.

20. “Draft a reply to the other party’s letter [letter text]: our position [position], a polite and firm tone, no admission of debt”.

21. “Compare the internal policy [text] with the requirements of [law and articles] and list the gaps”.

22. “Make a checklist for a junior lawyer to review a [type of contract]: the clauses that most often cause disputes, with an explanation”.

23. “Make a questionnaire for a first consultation on [type of case]: questions, documents, deadlines”.

24. “List the documents to request from [Company] for legal due diligence before the [type of deal]”.

25. “Check your previous answer: which laws and cases you named, whether each has full details, and what you are not sure about. List them for checking”.

Checking the answer

An AI answer is a draft. Before it goes to the client or the court, the lawyer checks four things.

1. Each law and each court decision exists, and the details and quote match the primary source.

2. The version of the law is in force on the right date.

3. The facts in the answer are in the source documents and were not added by the model.

4. The conclusion matches the client’s position and stays within the brief.

Stanford RegLab and HAI found errors in more than 17% of answers from Lexis+ AI and Ask Practical Law AI, and in more than 34% from Westlaw AI-Assisted Research. In an earlier study, general chatbots made errors on 58–82% of legal questions, so prompt 25 on the list does not replace a manual check.

What you must not paste into a prompt

Under Article 8 of Russia’s Federal Law on Advocacy, attorney-client privilege covers any information related to legal help for the client, and only the client can lift it.

On 17 February 2026, Judge Jed Rakoff (Southern District of New York) ruled in United States v. Heppner that the defendant’s chats with a public AI service were not protected by attorney-client privilege. The court cited the service’s policy, under which prompts and answers are collected and may be disclosed to third parties.

Before you send text to a public service, remove:

1. Names, passport details, addresses, and phone numbers of clients and their counterparties.

2. Case, contract, and invoice numbers, and amounts that could identify the client.

3. Information covered by attorney-client privilege or trade secrets, and the case strategy.

Replace them with labels such as [Client], [Counterparty], [amount], and keep the mapping table on your side. Work with full documents inside the firm’s systems or with a model on servers in Russia.

Sources and fact-checking

Checked by the editors on 16 October 2026.

  1. Stanford HAI, 23.05.2024 (hai.stanford.edu)
  2. Harvard Law Review Blog, 03.2026 (harvardlawreview.org)
  3. Great Russian Encyclopedia (bigenc.ru)

Found a mistake? Tell us

Publications

Mike Mozg

Agentic AI use cases at AT&T: nearly 500 processes now run on agents

AT&T no longer uses AI only to help staff work faster: it is rebuilding whole business processes around agents. SiliconANGLE reported on 14 November 2026 that nearly 500 agentic AI use cases at AT&T are already in production. Here is how the company rebuilds them, where a human still checks the agent, and what other large companies can take from it.

News4 min
Contact us