LLM prompting for lawyers: a practical guide with 30 examples

Lawyers are using large language models more than ever, but most aren't getting the results they need. The reason is almost always the same: the prompt. Structured prompting isn't a technical skill. It's a communication skill, and it's the difference between a useful answer and a confidently wrong one.

What is LLM prompting for lawyers?

LLM prompting for lawyers means giving AI tools structured instructions that define role, task, scope, jurisdiction, and output format. Good prompts reduce hallucination, improve consistency, and make AI-generated work easier to check.

An LLM, or large language model, is trained on huge volumes of text and generates responses based on patterns in that data, not by pulling from live legal databases. When you write to tools like Claude, ChatGPT, or Gemini, you're writing prompts: instructions telling the model what to do and how.

The stakes matter here. Courts have sanctioned lawyers for submitting AI-generated briefs with fabricated case citations. Bar associations across multiple jurisdictions have made clear that competence, confidentiality, and candor obligations apply fully to AI-assisted work.

Why structured prompts matter

Many lawyers treat AI like a search engine. That's the wrong mental model. Search engines retrieve existing documents. LLMs generate new text based on probability, which matters enormously in a field where precision and source accuracy aren't optional.

Hallucination risk is real. LLMs can produce plausible but entirely fabricated citations. A structured prompt that tells the model to flag uncertainty cuts this risk significantly.

Rules vary by jurisdiction. A prompt without a stated jurisdiction may answer under the wrong standard entirely.

Confidentiality obligations apply. Before sharing client documents with any AI tool, confirm your organization's data governance policy permits it.

Better prompts mean less review time. A structured, accurate output cuts down on the correction work that eats into whatever time AI was supposed to save.

SpotDraft's analysis of AI and law in 2025 shows teams adopting AI selectively, especially in research, contract management, and knowledge work. Structured prompting turns that interest into safe, repeatable execution.

The 4 core elements of an effective legal prompt

1. Define the role. Tell the model what kind of expert to act as. Example: "You are a senior in-house counsel specializing in commercial contracts."

2. Specify the task. Be precise. Vague tasks produce vague outputs. Example: "Review the attached NDA and flag clauses that deviate from our standard positions on confidentiality scope, exceptions, and return of information."

3. Constrain the scope. Limit sources, jurisdiction, and what the model shouldn't do. Example: "Base your analysis only on the attached document. Flag ambiguous provisions rather than interpreting them."

4. Define the output format. State exactly how you want results presented. Example: "List each clause with the reference, the issue, and a recommended revision."

Weak vs. strong prompt comparison

The strongest prompt defines role, task, source constraint, output format, and an uncertainty instruction. Each piece cuts the chance of a hallucinated or misaligned answer.

Common mistakes to avoid

Asking without context. "What's the standard notice period for termination?" is too generic without jurisdiction and contract type attached.

Skipping source constraints. Without telling the model to stick to the attached document, it may blend training data with the actual text.

Skipping jurisdiction. Always name the governing law, especially for enforceability or compliance questions.

Treating AI output as final. No output should reach a client, court, or counterparty without human review.

Sharing confidential information without clearance. Check your data governance policy before uploading anything sensitive.

Letting the model fill gaps. If a document doesn't state a term, ask the model to flag the gap instead of guessing.

A checklist for better legal AI prompts

• Did you assign the model a specific legal role?
• Did you describe the exact task clearly?
• Did you specify governing jurisdiction?
• Did you name and limit the model to your source material?
• Did you define the output format?
• Did you tell the model not to infer missing facts?
• Did you ask it to flag uncertainty instead of guessing?
• Did you confirm sharing this content is permitted?
• Have you planned for human review before using the output?

30 prompt examples by task type

Contract analysis and review

These work best when the contract text is pasted directly or attached. Pair these with a contract review checklist and tips for reviewing a contract faster for a fuller framework.

1. Contract summary: "You are a senior contracts attorney. Summarize this agreement based only on the attached text. Cover: (1) parties and purpose, (2) key obligations, (3) payment and renewal terms, (4) termination rights, (5) unusual clauses. Use bullet points. Flag ambiguous provisions."
2. Clause deviation review: "You are in-house counsel reviewing a vendor NDA. Compare it against these standard positions: [insert]. Note the clause reference, deviation, and a suggested revision for each. Don't infer intent."
3. Risk flagging: "You are a contracts attorney. Identify the top five highest-risk clauses for [party name]. Explain each risk in plain language with a mitigation approach."
4. Obligation extraction: "You are a contract analyst. Extract all obligations imposed on [party name]. Number each item with the clause reference, the obligation, and the deadline. Don't add obligations not explicitly stated."
5. Renewal and notice term extraction: "Extract all renewal, auto-renewal, and notice provisions. Present as a table: clause reference, summary, notice period, deadline."

Legal research

Always verify AI-generated research against primary sources. See SpotDraft's practical guidance on staying current with civil litigation for broader habits beyond prompting.

6. Issue spotting: "Based on this fact pattern, identify the top three legal issues a court applying [jurisdiction] law would consider. Explain the relevant standard for each. Flag uncertainty."
7. Jurisdiction comparison: "Compare how [jurisdiction A] and [jurisdiction B] treat [issue]. Summarize in a table. Note your knowledge cutoff and flag areas that may have changed."
8. Regulatory overview: "Provide an overview of compliance obligations under [regulation] for [company type]. List as numbered items and flag contested interpretations."
9. Statute summary: "Summarize [statute] as it applies to [context], in plain language for a non-lawyer audience, without changing the legal meaning."
10. Case law landscape: "Describe the general legal trend around [issue] in [jurisdiction] over the past 5 years. Don't cite specific cases unless certain they exist."

Policy and compliance

11. Policy gap analysis: "Review [policy] against [regulation]. Present as a table: requirement, current position, gap, recommended action."
12. Data privacy review: "Review this privacy policy against [applicable law]. Note the provision, the requirement it may miss, and a suggested revision."
13. AI policy assessment: "Review this AI use policy against [EU AI Act] requirements for [risk category]. Flag gaps in plain language."
14. Contract compliance check: "Confirm whether this agreement includes: [list]. Note present, absent, or partial for each. Don't add provisions not found in the document."
15. Vendor due diligence summary: "Summarize risk areas across data security, subprocessor use, liability limits, and business continuity from these vendor responses. Flag incomplete answers."

For privacy-heavy commercial work, compare these against best practices for drafting a data processing agreement.

Knowledge management

16. Playbook extraction: "Extract negotiation positions for [clauses] from this playbook. Present as: standard position, fallback, walk-away point."
17. Precedent summary: "Summarize this precedent agreement as a template reference. Note jurisdiction- or industry-specific clauses that may need adjustment."
18. FAQ generation: "Generate 10 likely employee FAQs from this policy, with plain-language answers drawn only from the text. Flag unanswered questions."
19. Clause library entry: "Draft a clause library entry for [clause type]: standard language, negotiation variables, common pushback, fallback positions."
20. Training scenario creation: "Create 3 training scenarios for junior staff from this contract, illustrating common negotiation challenges."

A prompt library works best paired with a system that preserves precedent and centralizes institutional knowledge. SpotDraft's piece on AI and legal knowledge management explores that model further.

Memo and document creation

All outputs need human review before use.

21. Legal memo draft: "Draft a memo to [audience] on [topic]: background, legal question, analysis under [jurisdiction], conclusion and recommendation. Flag unsettled areas."
22. Executive summary: "Prepare a 1-page summary of this agreement for a non-lawyer audience: parties, purpose, obligations, payment terms, termination, unusual provisions."
23. Redline rationale: "For each redline: [list], give a 1-sentence rationale suitable for sharing with the counterparty."
24. Client-facing explainer: "Explain this clause to a non-lawyer client in plain language: what it means, what it requires, what risks it poses. Flag anything the client should discuss with their attorney."
25. Escalation memo: "Draft a brief escalation memo to [stakeholder] covering [issue], the risk, and a recommended next step."

Since readability affects both negotiation speed and downstream interpretation, these work well paired with SpotDraft's principles for clear contract language.

Strategic and advisory

These require careful human review and shouldn't be treated as final legal advice.

26. Risk scenario planning: "Based on these terms [paste], identify three scenarios where [party] could face significant exposure. Give the trigger, likely claim, and mitigation for each."
27. Negotiation strategy brief: "Based on this agreement and [context], identify the top 3 negotiation priorities with an opening position and fallback for each."
28. Regulatory change impact: "Assess the impact of [regulatory change] on [company type]. Identify contracts or policies needing review. Present as a prioritized list."
29. M&A contract flag: "Identify provisions that could complicate an acquisition of [company], including change of control, assignment restrictions, and consent requirements."
30. Legal ops process improvement: "Based on this current review process [describe], identify 3 inefficiencies and one AI-assisted fix for each."

Strategic prompts work best connected to structured intake and negotiation workflows, like a deal desk process.

Building a prompt library for your team

A prompt library is a shared set of tested, approved prompts your team reuses across common tasks.

1. Start with your highest-volume tasks. Contract summaries, NDA reviews, policy gap analyses.
2. Test before publishing. Refine each prompt against real examples until results stay consistent.
3. Include usage guidance. Note the task, expected source material, and required human review for each entry.
4. Assign ownership. Set a legal ops owner and a review cadence every 6 months.
5. Integrate with your contract workflow. Prompts embedded directly into review and drafting workflows get used far more consistently than a document sitting in a shared drive.

Platforms that support AI-assisted contract management make this shift easier. See how AI contract review tools are changing legal workflows for a broader view of where this is heading.

For ChatGPT-specific prompt examples geared toward in-house teams, our ChatGPT prompts guide covers that narrower use case.

When not to use LLMs in legal work

Don't use them as a substitute for verified legal research, for final legal advice, or in time-critical situations without review. Don't share privileged information without clearance, and don't rely on general-purpose tools for highly specialized or unsettled legal questions. For guidance on choosing the right tool for the right job, see SpotDraft's approach to selecting and managing AI tools.

Bring structured prompting into your workflow

SpotDraft helps in-house legal teams move from ad hoc AI use to structured, workflow-integrated operations, with AI baked into contract review and legal intake rather than bolted on as a separate step. Book a demo to see how it works.