Turn vague legal requests into review-ready work with tested prompts for drafting, research, contracts, and due diligence, plus safeguards for human review.
The Best AI Prompts for Lawyers: Ready-to-Use Examples for Every Legal Task (2026) can turn a vague request into useful legal work. A commercial lawyer facing a tight signing deadline may ask AI to “review this agreement.” That prompt often produces a long summary, not a focused risk report. Better prompts give AI a role, task, context, limits, and review standard.
TL;DR
To be effective, a legal prompt should specify the task, the client’s position, the applicable jurisdiction, the source material, and the desired answer format.
Drafting, review, research, communication, due diligence, and document analysis are best handled with separate prompts.
Before offering conclusions, have AI surface missing facts, ambiguous language, underlying assumptions, and limitations in the available sources.
Every material legal output should include clause references, confidence levels, practical next steps, and a requirement for human review.
Tested prompts deliver greater value to legal teams when incorporated into playbooks and matter workflows.
In a single workspace, Lawxy brings together legal research, drafting, review, document analysis, and workflow automation.
What makes an AI prompt useful for lawyers?
A legal prompt should specify both the action the system must take and any limits on that action. Additionally, it should also define the matter at hand, the client’s position, and the result sought. Absent that context, the system may produce a polished response while missing the legal issue that actually requires analysis.
Use this structure to frame the prompt:
Role: Identify the legal professional or analyst the task requires.
Task: Describe one clear job.
Context: Add the facts, agreement type, parties, and business purpose.
Position: State which party you represent.
Rules: Set out the governing law, policy, playbook, or review standard.
Output: Define the required format, length, and level of detail.
Quality check: Include a check requiring the system to flag gaps, assumptions, and uncertainty.
A more focused version would read:
Act as commercial counsel for the customer. Review the attached software agreement under New York law. Identify terms that create unusual cost, data, renewal, liability, security, or termination risk. Rank each issue as high, medium, or low. Furthermore, quote the relevant language, explain the business impact, suggest replacement wording, and list any missing facts.
The second prompt narrows the task. It also gives the reviewer a clear way to check the answer. This is important: AI often sounds certain even when its analysis rests on incomplete facts.
In Formal Opinion 512, the American Bar Association addressed this concern. The opinion also discusses a lawyer’s duties when using generative AI, including competence, confidentiality, client communication, and reasonable review. A strong prompt supports those duties, but it does not replace them.
Use one prompt for one main goal whenever possible. A request to summarize, redraft, compare, research, and negotiate all at once can result in a shallow answer. Split complex work into stages. Ask AI to extract facts first, analyze the text second, and draft an answer third.
Before sending a prompt, check five points:
Did you remove unnecessary personal or confidential data?
Is the applicable law and relevant dates clear?
Have you specified the party whose interests should guide the analysis?
Did you ask for citations or references to specific clauses?
Did you explain the process a lawyer should use to review the result?
AI prompts for contract drafting
AI can help lawyers produce an initial draft more quickly. Additionally, the prompt works best when it includes the business deal, not only the document name. “Draft a vendor contract” leaves too many issues open. The system must guess the services, payment model, data access, risk split, and exit rights.
Start with the commercial facts. Then identify the clauses that need special treatment. Mark assumptions explicitly instead of silently filling gaps.
Draft a complete contract
Use this prompt when you need an initial contract for internal review:
Act as senior commercial counsel for [party]. Draft an agreement classified as [contract type] between [party] and [counterparty]. Business terms: - Services or goods: [describe them] - Price and payment: [insert terms] - Contract term: [insert term] - Territory: [insert territory] - Governing law: [insert jurisdiction] - Key risks: [list risks] Include definitions, scope, fees, performance duties, confidentiality, intellectual property, data protection, security, warranties, indemnity, liability limits, insurance, termination, dispute resolution, and boilerplate terms. Use plain commercial language. Protect [party] without making the draft unreasonable. Mark every assumption with “[ASSUMPTION].” Before drafting, list the information that is missing. After completing the draft, provide a concise risk summary.
The prompt gives AI a useful drafting frame and clearly defines the drafting parameters. Hidden assumptions are less likely to enter the agreement. Reviewers can concentrate on the remaining business decisions.
Draft defined terms
Defined terms can narrow or expand rights across an entire contract. Moreover, a seemingly minor change may affect payment, warranties, indemnity, or termination. Have AI trace those connections.
Review the attached agreement while representing [party]. Identify terms that need definitions or clearer scope. For each term, propose a definition consistent with the agreement’s language and business purpose. Avoid circular wording. Explain which clauses the definition affects. Flag any definition that could expand liability, diminish a remedy, or introduce uncertainty. Organize the analysis into three parts: the proposed definition, the affected clauses, and the drafting concern.
Compare each proposed definition against the agreement as a whole. That review can reveal conflicts between a definition and an operative clause.
Draft a limitation of liability clause
The liability clause needs to reflect the deal’s value as well as how the parties have allocated risk. AI may suggest language; evaluating the represented party’s commercial tolerance, however, requires specific instructions.
Draft, for [party], a limitation of liability clause in connection with a [contract type]. Set out the contract’s value as [amount] and its principal risks as follows: [risks]. Recommend a general liability cap based on [fees, fixed amount, or other measure]. Therefore, address direct damages, indirect damages, lost profits, data loss, confidentiality breaches, intellectual property claims, fraud, gross negligence, willful misconduct, payment duties, and indemnity obligations. Offer three positions, ranging from customer-friendly through balanced to supplier-friendly. Describe the business effect of each position. Note which issues may turn on [jurisdiction]. Do not state that a clause is enforceable without legal support.
This version gives the lawyer negotiation choices. It distinguishes the preferred client position from a market compromise.
Draft a negotiation fallback
A legal team may need several acceptable positions rather than just one. Ask AI to create a clear fallback ladder.
Based on the attached clause and our position, create a negotiation ladder with three options. Option one should reflect our preferred language. Option two should offer a reasonable compromise. For option three, state the least protection we can accept. Describe the legal and business tradeoffs associated with each option. Ensure the language is suitable for use in a redline. Exclude market data unless it is supported. Flag any decision that requires business approval.
A fallback ladder can help prevent rushed decisions during calls. The lawyer remains responsible for choosing the option appropriate to the matter.
AI prompts for contract review and redlining
Review prompts should focus on risk, not document summary. Additionally, a summary, by contrast, may give every section equal weight. Legal review should highlight what could harm the client, delay the deal, or create an unwanted duty.
The prompt should identify the review standard. That standard might come from a company playbook, prior agreement, regulatory rule, or client instruction.
Find risky clauses
Act as counsel for [party] and review [contract type] under [governing law]. Apply [governing law] to that agreement throughout your review. Treat the attached review playbook as the controlling standard. Identify clauses that: - Create unusual financial exposure - Shift operational duties to our client - Limit our rights or remedies - Permit unilateral changes - Create automatic renewal or difficult exit - Allow broad data use - Lack clear service or performance standards - Conflict with another clause For each finding, quote the clause, name the risk, rate its severity, explain the business impact, and suggest a practical revision. Where the contract lacks a requested protection, state “not found.”
This prompt creates a repeatable review record. Moreover, the “not found” instruction helps reveal missing terms instead of focusing only on existing text.
Check a contract against a playbook
Compare the attached agreement with the attached legal playbook. Review each playbook rule separately. For every rule, state whether the contract complies, partly complies, or does not comply. For each determination, cite the relevant contract section. Describe any difference in plain English. If you identify a gap, provide proposed replacement wording. Treat silence as noncompliance, and flag it as such. Also identify any playbook rule whose detail is insufficient for a reliable decision.
This approach suits procurement, sales, privacy, and vendor agreements. Furthermore, for junior lawyers, it provides a structured path through the review.
Prepare redline instructions
Review this agreement for [party]. Create redline instructions for counsel, not a final redline. Rank each change as required, preferred, or optional. Also, for every change, include the current language, the proposed concept, the reason, and the likely counterparty response. Focus on [issues]. Preserve business terms that do not affect those issues. Flag changes that need approval from security, finance, privacy, or the business owner.
The prompt keeps AI from rewriting the entire contract. It also separates legal judgment from drafting mechanics.
The National Institute of Standards and Technology provides the AI Risk Management Framework as a voluntary framework for managing AI risks. Its themes support a useful legal review habit: define the intended use, test the output, document limitations, and keep people accountable for decisions.
Summarize redlines for business leaders
Therefore, > Summarize the attached redline for a business executive. Use plain English and no legal jargon. Group changes into price, timing, duties, risk, data, termination, and approval needs. Explain what changed, why it matters, and what decision the business must make. Limit the summary to 500 words. For each material point, cite the applicable contract section. Distinguish agreed changes from open issues.
This prompt helps legal teams communicate without sending a clause-by-clause memo. It also keeps business leaders focused on decisions.
AI prompts for legal research and case analysis
Legal research requires more control than a general question and answer. Additionally, start by specifying the jurisdiction, court level, relevant time period, legal issue, procedural posture, and type of authority sought. Where reliable legal sources are available, have the tool provide citations and direct links.
Rather than asking AI to “find cases that support my argument,” provide enough detail to define the research task. That framing may encourage selective research. Ask for authorities that support the position, undermine it, or leave the issue uncertain.
Build a research plan
Serve as a legal research attorney. Begin with this issue: [issue]. The governing jurisdiction is [jurisdiction]. The relevant facts are: [facts]. The case is at the following procedural stage: [posture]. Identify the governing statute, controlling cases, persuasive cases, agency guidance, and unresolved questions in a research plan. Set out the order the research should follow. Include search terms and likely counterarguments. Do not draw a legal conclusion at this stage.
This approach can expose missing facts before a lawyer spends time searching. It also creates a record of the method used.
Find and assess authorities
Research whether [legal proposition] applies under [jurisdiction] as of [date]. Use primary sources where available. Provide the case name, court, date, citation, link, relevant facts, holding, and reasoning. Separate controlling authority from persuasive authority. Include cases that reach a different result. State whether each source directly supports the proposition, supports it by analogy, or cuts against it. Do not invent citations. Moreover, if you cannot verify a source, say so clearly.
This prompt treats source quality as part of the task. It also reduces the chance that a lawyer mistakes a related case for a controlling one.
Analyze a case record
Review the attached pleadings, orders, and evidence as counsel for [party]. Organize the case into a map covering material facts, disputed facts, claims, defenses, elements, evidence supporting each element, evidence gaps, procedural deadlines, and the principal counterarguments. For each factual statement, cite the supporting document and page. Separate proven facts, allegations, inferences, and unknown facts. Assess credibility only where the record supports doing so.
For large records, this format gives lawyers a structured way to manage the material. It also preserves the distinction between allegations and established facts.
Test an argument
Act as opposing counsel. Analyze our proposed argument: [argument]. Probe it for the strongest factual, legal, procedural, and evidentiary attacks. Cite the relevant record or authority. Then propose a revised argument that answers each attack. Mark any response that depends on a fact we have not confirmed. Keep the tone professional and direct.
A controlled opposition review can improve briefs, advice, and settlement strategy. It should challenge the argument without inventing facts.
The U.S. Courts website provides official information about federal courts, opinions, rules, and filing resources. Use official court sources where possible. Have AI supply source links, then open and verify each authority yourself.
AI prompts for client communication and legal operations
Legal teams spend many hours turning analysis into useful messages. Additionally, aI can support that work, so long as the prompt safeguards tone, accuracy, and privilege. Uncertainty should remain explicit in the output; the system must not recast it as a firm promise.
Draft a client email
To [client or business team], draft an email about [issue]. Cover the current position, its practical impact, available options, the deadline, and the recommended next step. Use plain English and a calm tone. Do not give a conclusion beyond the facts and authorities provided. Mark any missing fact with “[CONFIRM].” Keep the email under 350 words. End with three focused questions for the recipient.
This prompt gives the client a path forward. It also prevents the system from hiding open points inside polished prose.
Turn legal analysis into an executive brief
Convert the attached legal analysis into a one-page executive brief. Start with the decision needed. Then state the key facts, legal risk, business impact, options, recommendation, and next action. Use short headings and bullets. Avoid legal citations in the main text, but list the sources at the end. Preserve all stated limits and uncertainties. Do not add facts.
This output can support a board paper, deal meeting, or risk committee discussion. The lawyer should compare it with the source analysis before sending it.
Triage legal intake
Review this legal request and classify it as contract, dispute, privacy, employment, regulatory, intellectual property, or other. Extract the requester, business unit, deadline, affected entities, jurisdiction, documents, and requested outcome. Identify missing information. Assign a priority of urgent, high, normal, or low using these rules: [rules]. Draft five follow-up questions. Do not provide legal advice.
This prompt supports an intake team without turning intake into legal advice. It also creates consistent routing data.
The International Association of Privacy Professionals publishes practical resources on privacy programs and data governance. A legal intake workflow should follow the organization’s own privacy rules when it handles personal information. Limit the data placed into any AI system to what the task requires.
AI prompts for due diligence and document analysis
Due diligence often involves hundreds or thousands of documents. Additionally, aI can handle the initial pass by organizing materials, extracting key terms, and identifying gaps. It should not replace a lawyer’s review of high-risk documents.
Before analysis begins, specify the documents in scope and the questions to be answered. Include page references and state how confident you are in each finding. Set out how unusual or incomplete findings should be escalated.
Review a document set
Review the transaction materials relating to [target or asset]. Identify contracts, licenses, disputes, employment issues, intellectual property rights, data obligations, change-of-control clauses, termination rights, and unusual liabilities. Moreover, for each finding, provide the document name, page, issue, risk level, business impact, and follow-up question. Group duplicate findings. Mark missing documents separately. Treat the absence of an issue in the documents as inconclusive; do not infer that silence means no risk.
The resulting output gives the team a usable diligence tracker. It also distinguishes missing evidence from negative findings.
Extract obligations and dates
Extract every material obligation, renewal date, notice period, payment duty, service level, audit right, insurance requirement, and termination trigger from these documents. Furthermore, use the exact clause reference. Identify the responsible party, due date or trigger, consequence of failure, and source document. Flag dates that require calculation. Do not calculate a deadline unless the notice date and governing rule are clear.
This task can support post-signature contract management. A team should verify the deadline through human review before relying on it.
Ask questions across documents
Using only the materials provided, answer this question: [question]. Also, cite the source document name and page for every answer. If documents conflict, identify the conflict. If the documents do not answer the question, state “not answered in the provided documents.” Separate direct statements from reasonable inferences. Do not use outside knowledge.
This approach reduces the risk of unsupported answers. It also signals clearly when the available materials do not contain the needed information.
Create a diligence report
Prepare a diligence report for [audience]. Therefore, organize findings under commercial, legal, financial, regulatory, privacy, employment, intellectual property, and dispute risk. Therefore, rank issues by severity and deal impact. Include a short executive summary, detailed findings, missing documents, recommended actions, and questions for management. Cite each finding. Make clear which points are verified facts, which come from management, and which remain open issues.
Its structure should reflect the transaction. A buyer’s report could include suggestions for allocating risk. By contrast, an internal legal team may need a closing checklist.
Adapt the report’s structure to the transaction. A buyer’s version may include risk-allocation suggestions. An internal legal team may instead need a closing checklist.
How should lawyers validate AI output?
AI output needs a review process that matches the legal risk. Additionally, a typo in an internal summary poses a different risk from a false case citation in a court filing. Use stronger controls for higher-impact work.
Start with source validation. Open every cited case, statute, clause, and page reference. Check whether the source supports the exact statement. Confirm that the authority remains current and applies to the right jurisdiction.
Then check the reasoning. Ask whether AI:
Used the correct party position
Applied the correct governing law
Distinguished facts from assumptions
Found adverse authorities
Missed exceptions or definitions
Read related clauses together
Preserved the client’s instructions
Suggested action without authority to do so
Use a second prompt as a quality check:
Audit the answer above. List every factual claim, legal claim, assumption, and recommendation. For each item, state the supporting source or write “unsupported.” Identify any missing adverse authority, unclear fact, incorrect citation, or conclusion that exceeds the evidence. Do not rewrite the answer during this stage.
An audit prompt, on its own, cannot establish the answer’s accuracy. It can nevertheless give the lawyer a faster review list. Responsibility for the final analysis remains with the lawyer.
Create a simple approval process for recurring work:
Define approved use cases.
Set data handling rules.
Tested prompts should be maintained in a controlled library.
Legal claims should undergo source checks.
For high-risk outputs, assign review to a qualified lawyer.
The record should document any material edits and the decisions approving the output.
Prompts should be reviewed whenever the law or a relevant policy changes.
The European Commission’s AI Act materials discuss risk-based obligations for AI systems in the European Union. Organizations should assess whether their legal AI use creates additional compliance duties. They should also align AI use with confidentiality, privilege, information security, and records policies.
Prompt libraries need maintenance. Performance can vary by contract type, even when a prompt has worked well elsewhere. For each saved prompt, record the document type, jurisdiction, audience, and review date. When errors recur, track them and revise the prompt accordingly.
How legal AI software solves this
Generic AI tools can assist with drafting, review, research, document questions, and summaries. Additionally, purpose-built legal software, in turn, adds structured workflows, source grounding, playbooks, document controls, and human approvals. Lawxy brings these functions together through contract review, legal research, document intelligence, due diligence, drafting, and workflow automation.
Lawxy can support clause-level review, citation-backed research, cross-document questions, bulk diligence, Microsoft Word work, and AI agents with human approval steps. These features help teams reuse sound prompts while keeping important decisions with people.
See how Lawxy Legal AI Software helps legal teams work with confidence.
FAQ
Can lawyers use AI to draft legal documents?
Yes. Additionally, aI can create a first draft, suggest clauses, and improve structure. A qualified lawyer must check the facts, law, client position, and final language.
What should every legal AI prompt include?
Include the task, facts, jurisdiction, client position, source material, output format, and review rules. Have it flag missing facts and unsupported conclusions.
How can lawyers reduce hallucinated citations?
Moreover, require primary sources, direct links, quotations, and page references. Verify every citation in the original source before relying on it.
Should lawyers include confidential information in AI prompts?
Only if the tool, account, and workflow meet the organization’s confidentiality and security rules. Remove personal and sensitive data when the task does not need it.
Can AI assess the acceptability of a contract term?
Furthermore, can lawyers rely on AI to determine whether a contract provision is acceptable?
The tool can identify risks and compare the language with a playbook. That decision belongs to the lawyer and business owner, who must determine whether the risk fits the deal.
How can legal teams build a prompt library?
Group prompts by task, document type, jurisdiction, and audience. Test each prompt, record known limits, set an owner, and review it after legal or policy changes.
Also, what is a good prompt for contract review?
Tell AI to act for a named party, follow a defined playbook, cite each clause, rank risks, explain business impact, and suggest practical revisions. Require it to identify missing protections.
Can AI research current law?
Only when the tool can access current, reliable legal sources. Ask for source dates and links, then verify each authority independently.
How should lawyers review AI-generated client emails?
Compare the email with the approved legal analysis and matter facts. Check tone, confidentiality, deadlines, legal certainty, and any statement that could create a commitment.
