ChatGPT Prompts for Lawyers: 20 Prompts Tested in Real Legal Work

ChatGPT prompts for lawyers are reusable instructions that turn a general-purpose model into a first-draft engine for one legal task. The twenty prompts below start from a fill-in template and then cover contract review, legal research, document drafting, client correspondence, and practice administration. Two further sets give in-house legal teams prompts for explaining a contract to a business client and for verifying what ChatGPT returned. One template underlies every ChatGPT prompt for lawyers below, stating the role, the client type, the governing jurisdiction, the facts to paste, and the output format. Every one returns a first draft that still requires verification before it leaves the building.

Key Takeaways

1. Effective ChatGPT prompts for lawyers set five variables: the role, the client type, the governing jurisdiction, the facts, and the output format. The template in the next section sets all five.
2. Redacting names, deal values and identifiers before pasting is the default, and every prompt that turns on jurisdiction states the governing law explicitly.
3. Effective ChatGPT prompts for lawyers still return drafts that need independent verification against primary sources.

[cta-1]

A ChatGPT Prompt Template for Lawyers

A ChatGPT legal prompt generator is most useful as a fill-in template, and the ChatGPT prompt template below is the one every prompt in this library uses. The block sets all five variables, and it is the starting point for any legal task the twenty prompts below do not already cover.

Act as a [practice area] lawyer advising a [client type] under [governing law]. Here are the facts, with names removed: [paste]. Produce [output format]. Flag every assumption you made, every fact you are missing, and every statement I should verify against primary sources.

1. [practice area] sets the vocabulary and the default conventions, so "commercial contracts" and "employment" return different redlines on the same clause.
2. [client type] decides whether the answer is a memo for a partner or a two-line note for a procurement lead.
3. [governing law] decides which rule the model applies. Left blank, it answers from whichever jurisdiction is best represented in its training data.
4. [paste] is the facts or the document text, with names, deal values and identifiers removed first.
5. [output format] is the table, memo, checklist or email the work product has to be, stated before the model picks one.

Why each element changes the output, and how a thin prompt differs from a specified one, is the prompt anatomy covered in our guide to writing a legal prompt for ChatGPT.

ChatGPT Prompts for Reviewing a Contract

A ChatGPT prompt to review a contract works when it names the agreement type, the governing law, and the side the lawyer is on. The four prompts below do that for a first-pass summary, a risk triage, a negotiation read, and a comparison against a team template.

Prompt 1: The Complete Contract Overview (Used by Transactional Lawyers)

Analyze this [contract type] governed by [jurisdiction] law. Identify 'killer' clauses and check for consistency with the term sheet.

Provide a comprehensive summary organized as follows:

1. Parties and their roles
2. Key business terms (scope, price, duration)
3. Material obligations for each party
4. Termination rights and conditions
5. Liability allocation and caps
6. Notable provisions (unusual terms, missing standard clauses)
7. Potential risks or red flags

Format as a table with columns for: Provision Category, Details, and Risk Level (Low/Medium/High).

What this does: The output becomes the first column of the issues list when a contract arrives from opposing counsel. The recurring failure is the model reporting a liability cap as present when the clause only cross-references one, so the caps row gets checked against the document text.

Prompt 2: The Risk Spotter (Used by In-House Counsel)

You are in-house counsel for [company type/industry]. Review this [type of contract] and identify all provisions that create:

1. Financial liability or exposure above $[amount]
2. Operational restrictions that could limit business flexibility
3. Indemnification obligations (especially one-sided)
4. IP ownership or licensing concerns
5. Compliance or regulatory risks
6. Termination vulnerabilities

For each risk, explain the business impact and suggest mitigation language.

What this does: This sorts a review queue into what needs negotiation and what can be signed. The dollar threshold decides whether it works: set too low, every agreement comes back flagged and the triage has achieved nothing.

Prompt 3: The Negotiation Position Analyzer (Used by Deal Lawyers)

Analyze this [type of contract] to assess negotiation leverage:

1. Which provisions heavily favor our side vs. the counterparty?
2. What are the top 3 negotiation priorities based on risk exposure?
3. Which terms are likely deal-breakers vs. negotiable?
4. What alternative language could we propose for problematic clauses?
5. Where does the other party have leverage we need to acknowledge?

Present findings as a negotiation roadmap with "Must Win," "Should Win," and "Can Concede" categories.

What this does: Deal lawyers run this before a negotiation call. The output is a hypothesis, not a position: the model has no view of the commercial relationship or the pressure to close, so the "Can Concede" column is rewritten first.

Prompt 4: The Contract Gap Analysis (Used by In-House Legal Teams)

Compare this [contract type] against our attached '[contract type] Template' for [industry/deal type]. Identify:

1. Missing provisions
2. Provisions present but inadequately detailed
3. Absent protections
4. Clauses that deviate from the template

Recommend specific additions or modifications to align the contract with our standard template.

What this does: In-house counsel run this before approving a contract, which can help ensure the protections the team standardized on are present. It is most useful on vendor-drafted agreements that omit client-favorable terms. Practitioners add a second pass asking which flagged deviations the team already accepts.

Legal Research Prompts

A ChatGPT prompt for legal research produces a research map, not an authority. ChatGPT does not carry the citator, coverage or currency guarantees of a legal research database, and the two prompts below are written to organize the question before a lawyer opens one.

Prompt 5: The Issue Research Framework (Used by Litigation Associates)

Research [issue] under [Specific State] Revised Statutes Chapter [X].

Key facts: [2-3 sentence summary]

Provide:

1. Applicable legal standard and source
2. The majority view and supporting authority
3. Minority positions (if any)
4. Jurisdictional splits or circuit conflicts
5. Analogous cases with similar fact patterns
6. Practical application to these facts
7. Potential counterarguments

Flag any unsettled areas or conflicting authority.

What this does: The output names candidate standards and cases, which a lawyer confirms in a reporter. The fabrication pattern in legal AI output is why every item is treated as a search term rather than authority.

Prompt 6: The Memo First Draft Generator (Used by Law Firm Associates)

Draft a legal memorandum analyzing: [issue]

Format: IRAC (Issue, Rule, Analysis, Conclusion)

Jurisdiction: [specify]

Key facts: [brief summary]

Applicable law: [statutes/case law if known]

Structure the memo with:

1. Brief issue statement
2. Governing legal standard with citations
3. Application of law to facts (step-by-step analysis)
4. Counterarguments and responses
5. Conclusion with confidence level (strong/moderate/weak)

Write in formal legal memo style, approximately [word count].

What this does: This returns a memo skeleton an associate fills with verified research. The confidence line is what goes wrong: the model states "strong" on unverified authority, so the conclusion is rewritten last, after the citations are checked.

Document Drafting Prompts

Using GPT for legal documents works best on the sections that follow a fixed form, such as a statement of a legal standard or a set of discovery requests. The two prompts below cover both, and each needs the governing rule named in the prompt.

Prompt 7: The Legal Standard Framework Prompt (Used by Litigators)

Act as a senior litigation associate. Draft the 'Legal Standard' section for a [Motion Type, e.g., Motion to Dismiss for Failure to State a Claim] pursuant to [Specific Rule, e.g., FRCP 12(b)(6)].

Jurisdiction: [Specify Court, e.g., Southern District of New York]

Governing Authority: Focus on the standard established by [e.g., Twombly and Iqbal] and relevant circuit-specific precedents regarding the 'plausibility' requirement.

Requirements:

1. Provide a concise, neutral statement of the burden of proof.
2. Include the standard for how the court must treat well-pleaded factual allegations vs. legal conclusions.
3. Format citations in [Bluebook/Local Rule] style.
4. Do not draft a 'Statement of Facts' or 'Argument' section. Do not attempt to apply the law to specific facts. Provide only the governing legal framework.

What this does: The legal standard section is the most formulaic part of a brief, so a first draft frees the time for the argument. The dismissal standard comes from Federal Rule of Civil Procedure 12, and the circuit gloss on plausibility varies, so the draft is checked against the forum court before filing.

Prompt 8: The Discovery Request Builder (Used by Litigation Teams)

Draft [interrogatories/requests for production/requests for admission] for a [case type] lawsuit.

Case theory: [brief explanation]

Information needed: [specific categories]

Legal Standard: All requests must comply with the scope defined in FRCP Rule 26(b)(1). Every request must be proportional to the needs of the case, considering the importance of the issues at stake and whether the burden of the discovery outweighs its likely benefit.

Create [number] requests that:

1. Are specific enough to be enforceable
2. Cover all relevant information categories
3. Comply with proportionality standards
4. Avoid overly broad or harassing requests
5. Follow [jurisdiction]'s discovery rules

Organize by topic with clear instructions to the responding party.

What this does: The output is a coverage draft, broad enough that nothing in the case theory is missed. Proportionality is defined in Federal Rule of Civil Procedure 26, and a model drafting to it produces too many requests, so the editing pass cuts rather than adds.

Client Communication Prompts

A ChatGPT prompt for a client update email handles routine correspondence, along with the professional-response prompt below it. The tone decisions, and the decision about what to tell the client at all, stay with the lawyer.

Prompt 9: The Client Update Email (Used Across All Practice Areas)

Draft a client status update email for [client type: sophisticated GC/individual client/small business owner].

Tone: [professional and reassuring/formal/casual-professional]

Include:

1. [Recent development 1]
2. [Recent development 2]
3. Next steps and timeline
4. Action items (if any) for client
5. Offer to discuss via call

Length: [3-4 paragraphs/brief/detailed]

Close with: [invitation to discuss/confirmation of next meeting/etc.]

What this does: The output arrives at the right length and register for the relationship named in the brackets. The practitioner edit is the lead sentence: the model reports neutrally when the client needs telling plainly that a deadline moved.

Prompt 10: The Professional Response Template (Used Across All Practice Areas)

Draft a professional response to: [email/letter/request]

Recipient: [opposing counsel/client/court/third party]

Key points to address: [list]

Tone: [firm but diplomatic/neutral and factual/cordial]

Purpose: [inform/decline/accept/request extension/etc.]

Length: [brief/moderate/detailed]

Include any necessary caveats about [privilege/without prejudice/etc.]

What this does: This covers the routine inbox so substantive matters keep the attention. The caveat line is what to watch: a model placing a without-prejudice legend sometimes characterizes the underlying position, which is how a response meant to decline an extension concedes a fact.

Practice Administration Prompts

A ChatGPT prompt to extract deadlines and obligations from a contract turns a signed agreement into a dated checklist. The three prompts below cover the administrative work that consumes legal time without requiring legal analysis: meeting notes, compliance checklists, and obligation tracking.

Prompt 11: The Meeting Summary Generator (Used Across All Practice Areas)

Convert these meeting notes into a professional summary:

[Paste raw notes or bullet points]

Format for: [internal file memo/client summary/team distribution]

Include:

1. Meeting date and participants
2. Key topics discussed
3. Decisions made
4. Action items (assign to specific people with deadlines)
5. Next steps

Tone: [professional/formal/internal-casual]

What this does: Lawyers run this immediately after a call, while the detail is still recoverable from rough notes. The check is which action items came back without an owner, because those are the ones nobody has.

Prompt 12: The Compliance Checklist Generator (Used by Regulatory and Compliance Lawyers)

Create a comprehensive compliance checklist for [regulation/requirement].

Context: [company type/industry/jurisdiction]

Scope: [full compliance audit/initial assessment/specific area]

Include:

1. Regulatory requirements organized by category
2. Documentation needed for each requirement
3. Responsible parties/departments
4. Completion deadlines (if applicable)
5. Risk level for non-compliance (High/Medium/Low)

Format as a table with a checkbox column for tracking.

What this does: Compliance counsel use this when a new requirement takes effect and a first checklist has to exist. The output is organized by requirement, which is not how the work gets done, so the practitioner step is re-cutting the table by owning department and confirming each entry against the regulator's published text.

Prompt 13: The Deadline and Obligation Tracker (Used by Litigation and Transactional Teams)

Extract all deadlines, obligations, and key dates from this [contract/court order/settlement agreement].

Create an organized list including:

1. All specific deadlines with dates
2. Ongoing obligations and their frequency
3. Contingent obligations and triggering events
4. Notice requirements and timing
5. Action items with responsible parties

Format as a chronological timeline with categorization by obligation type.

What this does: The output becomes the matter calendar. Contingent obligations are where it fails: the model lists a triggering event and omits the notice period that runs from it, so the notice column is reconciled against the clause text. A team running this across a portfolio is better served keeping prompts in a shared library than rebuilding the instruction.

Advanced ChatGPT Prompts for Lawyers

Advanced ChatGPT prompts for lawyers ask the model to hold more than one position at once: several contracts, several perspectives, or several risk allocations of the same clause. Once the basics are familiar, more advanced prompting supports deeper analysis.

Prompt 14: The Multi-Contract Comparison (Used by In-House Counsel)

Compare these [number] contracts for [contract type]:

[Paste or upload contracts]

Analyze and report:

1. Standard terms that appear across all contracts
2. Provisions that vary between contracts (highlight deviations)
3. Most favorable terms for [our company], and which contract contains them
4. Least favorable terms and suggested improvements
5. Missing provisions that should be standardized
6. Recommendations for template language

Present as a comparison matrix with findings summary.

What this does: In-house counsel use this when rationalizing a vendor portfolio or running M&A due diligence on a target's contracts. Context limits bind: a model given a dozen agreements summarizes the first few and generalizes the rest, so the comparison runs in small batches.

Prompt 15: The Strategic Issue Analysis (Used by Senior Associates)

Analyze this situation from three perspectives:

[Provide fact pattern]

1. As counsel for [Client]: Identify strongest claims, defenses, and strategic advantages
2. As opposing counsel: Identify our weaknesses and their best arguments
3. As the judge: Assess the likely outcome and evaluate both sides' positions

For each perspective, provide:

1. Key legal and factual issues
2. Strongest arguments
3. Vulnerabilities
4. Strategic considerations
5. Recommended approach

Conclude with an overall risk assessment and strategy recommendation.

What this does: This surfaces arguments a lawyer has not reached yet, particularly in the opposing-counsel role, where the model is less invested in the client's theory than the person drafting is. The outcome prediction in the third role is the weakest part and is read as a prompt for further thought.

Prompt 16: The Clause Library Builder (Used by Transactional Teams)

Generate [number] variations of a [clause type] for [contract type] governed by [jurisdiction] law.

Create versions with different risk allocations:

1. Client-favorable (maximum protection)
2. Balanced/market-standard
3. Counterparty-favorable (used when we have limited leverage)

For each variation:

1. Explain the risk allocation
2. Note key protective language
3. Identify when to use this version

Include optional add-ons for special situations.

What this does: Transactional teams generate starting-point options, then narrow them against positions the team has actually held. The "balanced/market-standard" label deserves suspicion: a general-purpose model has no measured market data behind it. The same structures written without reference to a specific model sit in the wider collection of prompts written for legal work.

[cta-1]

Prompts for Explaining a Contract to a Business Client

A ChatGPT prompt to explain a contract clause in plain English is what lets a business approver act on it without reading the agreement. The two prompts below turn a ChatGPT-reviewed clause into a plain-English approval request a deal lead can answer.

Prompt 17: The Plain-English Clause Translation (Used by Commercial Counsel)

Explain the following clause from a [contract type] to a non-lawyer [role, e.g., procurement lead] who has to approve it.

Clause text: [paste]

Produce:

1. One sentence on what the clause requires, in plain language
2. What it means for us in practice (money, time, or restriction)
3. Which terms in it are one-sided, and in whose favor
4. The one question the approver should ask before signing

Do not use defined terms from the contract without explaining them. Do not give a recommendation.

What this does: This turns a flagged clause into an approval request a business owner can answer. Withholding the recommendation is deliberate, because that call is the lawyer's. The output is checked against the clause text for dropped conditions, since plain-English rewrites lose carve-outs first.

Prompt 18: The Deal Risk Note (Used by In-House Legal Teams)

Turn the completed review below into a short risk note for the [deal lead/business owner] on this transaction.

Review findings: [paste]

Produce:

1. Three sentences on the overall risk posture of the agreement
2. The three issues that need a decision, each with the business consequence
3. The issues already resolved, listed without commentary
4. What we are asking the counterparty for, in one list

Length: under 300 words. No legal citations.

What this does: This converts a clause-level review into the document the business actually reads, which decides whether legal findings change the deal. The practitioner edit is the ordering: a model ranks by legal severity, and a deal lead needs the issues ranked by what holds up signature.

Prompts for Verifying ChatGPT Legal Output

A ChatGPT prompt to verify ChatGPT legal output is the last step in every workflow on this page. A verification prompt surfaces some fabrication and not all of it, so the lawyer still reads the authority. The two prompts below narrow where to look.

Prompt 19: The Source Separation Check (Used Across All Practice Areas)

Review your previous response. Separate it into two lists:

1. Statements supported by text I provided, with the quoted supporting language for each
2. Statements you supplied from general knowledge, training data, or inference

For list 2, mark each statement as: a legal standard, a case citation, a market norm, or an assumption about the facts.

Do not add new analysis. Do not defend the previous response.

What this does: This separates what the model read from what it supplied, and anything in the second list is unverified by definition. The limit is real: a model auditing itself will sometimes place a fabricated citation in the first list, which is how fabricated citations reached a court, so the quoted language is checked against the document.

Prompt 20: The Verification Checklist (Used Before Anything Leaves the Office)

List every citation, legal standard, statutory reference, and factual assertion in your previous response as a verification checklist.

For each item, give:

1. The exact claim as stated
2. The type of source needed to confirm it (reporter, statute, regulation, contract text, client confirmation)
3. Where in the response it appears
4. How the conclusion changes if the item turns out to be wrong

Order the list by how much the conclusion depends on the item.

What this does: The output is the checklist a lawyer works through before a draft leaves the office, ordered so load-bearing items are checked first. The duty is not discretionary: Federal Rule of Civil Procedure 11 makes the signing attorney responsible for the legal contentions in a filing.

[cta-2]

What Makes a ChatGPT Legal Prompt Work

What makes a ChatGPT legal prompt work is how much of the task the lawyer specifies and how much is left to the model. The five rules below cover what a prompt has to state, what it should leave out, and what to do when the first answer is close.

1. State the governing law. Without it the model answers from whichever jurisdiction is best represented in its training data, which is how a Delaware-law question comes back with California authority.
2. Name the output format before the model picks one. "Draft a contract" returns prose. "Draft a mutual nondisclosure agreement for two software companies under Delaware law, with a two-year confidentiality term and standard carve-outs" returns a usable first draft.
3. Refine instead of restarting. A near-miss is usually two clauses from working, and a correction on the same thread beats a second attempt from a blank prompt.
4. Keep the strategic call out of the prompt. The model can list the indemnity positions on the table. Whether to concede one of them is a judgment that stays with the lawyer who signs the advice.
5. Specify the posture, not just the task. The same clause text produces different recommendations for a party with leverage and a party without it, and the model cannot know which applies.

Data Handling and Professional Conduct Rules

What a lawyer may paste into a general-purpose model, and on which plan, is a data-handling question covered in the separate note on whether ChatGPT keeps what lawyers type into it, which treats the plan tiers, the retention controls, and the redaction habit that goes with them.

Whether the use is permissible in a given jurisdiction, and what a court or client has to be told about it, turns on professional conduct rules rather than on how the prompt is written. That question has its own treatment of lawyers working with ChatGPT, including the standing orders some courts now apply to AI-assisted filings.

Where Spellbook Fits

Spellbook is legal AI that reviews contracts without prompt engineering. It is the AI suite for transactional legal work, so a lawyer opens an agreement and asks a question in plain language rather than constructing a prompt around it.

1. The Review feature surfaces contract risks and proposes redlines as native Word tracked changes, so the output lands as an edit in the document rather than text to copy across.
2. The Draft feature completes clauses and whole sections using GPT-5, grounded in the team's own playbooks and precedent rather than in general training data.
3. Playbooks hold the positions a team has agreed to take, which is what makes a review consistent across every lawyer who runs it.
4. Compare to Market benchmarks a term against 20M+ real contracts and 2,300+ benchmark standards, which a general-purpose model cannot do, having no measured market set behind it.
5. Works in Word and Google Docs, so the review happens where drafting already happens.

On data handling, the suite operates under zero data retention, holds SOC 2 Type II, and does not use customer contracts to train public models.

Spellbook is a commercial-contract specialist rather than a legal-research platform or a litigation tool, so the research and motion-drafting prompts above still belong in ChatGPT and in Westlaw. On contract work, prompt engineering stops being a skill a lawyer has to acquire. Try Spellbook free with a seven-day trial.

Frequently Asked Questions

Can I Trust ChatGPT's Legal Research and Case Citations?

No. ChatGPT can hallucinate cases, fabricate citations, and misstate holdings while sounding completely confident. Every citation must be independently verified in Westlaw, Lexis, or an official reporter before it is relied on. ChatGPT is useful for research organization and initial analysis, never as a sole source.

What Are Some Effective AI Prompts for Lawyers?

The effective ones aim at a bounded task with a fixed output form: a clause-by-clause contract summary, a deadline extraction, an IRAC memo skeleton, a client status email, a plain-English translation of one provision. Prompts asking for a strategic recommendation, a settlement figure, or a prediction of how a court will rule return confident prose a lawyer cannot use.

What Is the Best Prompt Generator for ChatGPT?

For legal work, none of them. A general-purpose generator has to guess the variables that decide the answer: the governing law, the procedural posture, the counterparty, and the work product the output has to become. A template the lawyer completes is likely to outperform one that guesses, which is why the section above is a fill-in block rather than a tool.

Do ChatGPT Legal Prompts Work in Gemini or Copilot?

Mostly, with one adjustment. The five variables transfer unchanged, because they describe the task rather than the model. What differs is how much of a pasted agreement each model retains, so a prompt that works on a 10-page contract in one may need to run section by section in another. For model-specific phrasing, see prompts tailored for Gemini.