ChatGPT Prompts for Lawyers (60 Templates for 2026)
A litigation partner at a mid-size Chicago firm cut discovery-response drafting from six hours to ninety minutes last quarter. She did not add associates. She did not buy a $50k-per-seat legal AI platform. She rewrote her ChatGPT prompts. That is the entire delta — better inputs, faster outputs, same billable quality. The 60 templates below are the ones working in real practices right now, from solo shops to AmLaw 200 teams. Each is scoped to a specific task, written in the voice attorneys actually use, and paired with the ethics guardrails you need before you paste anything into a chat window.
Table of Contents
What Attorneys Actually Use ChatGPT For in 2026
Two years past the initial wave, the usage picture has stabilized. Thomson Reuters’ 2025 Future of Professionals survey found roughly a quarter of law firms had integrated generative AI into some part of their workflow, with adoption skewing higher inside larger firms.
Solo and small-firm patterns. Solos lean on ChatGPT for the work that used to eat evenings: intake summaries, first-draft engagement letters, plain-English client updates, demand-letter skeletons, statute translation. Every hour on non-billable drafting is an hour of margin gone.
Mid-size and BigLaw patterns. Larger firms use it for deposition prep outlines, deposition summary reformatting, timeline construction, first-pass redlines against a house playbook, and internal training material. The workflow is "sanitize, then prompt, then verify."
What almost nobody does well yet. Case research from scratch. Jurisdictional analysis. Anything requiring a verified citation trail. Those tasks belong in Lexis+ AI, Westlaw Precision AI, or a legal-specialized tool tied to real case databases — the single biggest source of malpractice risk in the room.
Ethics Rules Before You Prompt (Mandatory Read)
Before a single prompt, three ABA Model Rules govern almost every AI decision in a US law practice.
Model Rule 1.1 — Competence. Comment 8 requires lawyers to keep abreast of "the benefits and risks associated with relevant technology." As of 2024, roughly forty state bars had adopted some version of the duty of technological competence. If you use ChatGPT for client work, you are expected to understand what it does, where it fails, and how to verify its output.
Model Rule 1.6 — Confidentiality. Pasting client-identifying facts into a consumer ChatGPT account is a Rule 1.6 problem. Free and standard ChatGPT Plus tiers may use inputs for training unless you toggle that off. ChatGPT Team, Enterprise, and the API do not train on your data by default — that is the distinction that matters. Use those tiers or a legal-specialized tool with a signed BAA/DPA when client facts are involved.
Model Rule 3.3 — Candor to the Tribunal. This is where Mata v. Avianca earned its permanent spot in every AI-and-law CLE deck. In 2023, attorney Steven Schwartz filed a brief in the Southern District of New York containing six citations to cases that did not exist — ChatGPT had hallucinated them. Judge P. Kevin Castel sanctioned Schwartz and his co-counsel. If you file AI-generated content without independent verification, you own every fabricated cite.
The pre-prompt checklist. (1) Am I on a data-protected tier? (2) Have I redacted or fictionalized identifying facts? (3) Do I have a verification plan for every factual claim in the output? If any answer is no, close the tab.
15 Contract Review & Drafting Prompts
These prompts assume you are working with fictionalized or redacted text on a data-protected tier. Paste your contract, then the prompt.
1. Playbook-Based Redline Review
Act as senior transactional counsel. Review the pasted [contract type] against these house playbook positions: [paste 5-10 rules]. For each deviation, output a three-column table: Clause & Section | Deviation | Suggested Redline. Flag material commercial or liability impact as HIGH. Skip stylistic issues.
2. Indemnification Risk Heatmap
Analyze the indemnification and limitation-of-liability sections in the pasted agreement. Identify (a) mutual vs one-way indemnities, (b) carve-outs from the liability cap, (c) any uncapped exposure. Output a Low / Medium / High heatmap with one-sentence rationale each. Assume I am counsel for the [buyer / seller / licensee].
3. Change-of-Control & Assignment Sweep
Scan the pasted agreement for every clause touching assignment, change of control, or successor liability. Quote the exact language, cite the section, and note whether consent is required, deemed given, or silent. Flag anti-assignment provisions that would trigger in a private-equity roll-up.
4. SaaS MSA First Draft
Draft a Master Services Agreement for a US-based B2B SaaS vendor selling to enterprise customers. Include: subscription terms, SLA with 99.9% uptime and service credits, DPA reference, IP ownership, liability cap at 12 months fees, mutual IP-infringement indemnity, 30-day termination for convenience. Governing law: Delaware. No unrequested boilerplate.
5. NDA Comparison & Merge
I will paste two NDAs (mine and counterparty’s). Produce a clause-by-clause comparison table, then draft a merged version resolving conflicts in favor of the stricter confidentiality obligation and the more mutual position on remedies. Flag any clause where a reasonable negotiator would push back.
6. Employment Agreement — State-Specific Flag
Review the pasted employment agreement. Flag any non-compete, non-solicit, or garden-leave provisions unenforceable or narrowly construed in [state]. Cite the general legal principle, not specific cases — I will verify authority. Suggest replacement language that would survive scrutiny in that jurisdiction.
7. Force Majeure Modernization
Rewrite the pasted force majeure clause to expressly address pandemics, cyber incidents, government-mandated shutdowns, subcontractor supply-chain failures, and generative-AI-related service disruptions. Preserve the existing notice and mitigation obligations.
8. Data Processing Addendum Gap Check
Compare the pasted DPA against GDPR Article 28 and CCPA/CPRA service provider requirements. List gaps: Requirement | Present / Missing / Partial. Suggest replacement language for anything Missing or Partial. Scope is contractual only.
9. Contract Summary for a Business Client
Summarize the pasted [contract type] in under 400 words for a non-lawyer CEO. Cover: what we’re paying, what we’re getting, how we get out, what could go wrong. Plain English. No legalese. No "notwithstanding."
10. Renewal & Auto-Renewal Audit
Extract every renewal, auto-renewal, and termination-notice clause. Output a table: Clause | Section | Notice Window | Effective Date | Consequence if Missed. Add a calendar-ready summary line I can forward to my client’s ops team.
11. LOI to Definitive Agreement Bridge
I will paste a signed asset-purchase LOI. Draft the deal-point framework for the definitive APA: purchase price mechanics, escrow, reps & warranties survival, indemnity baskets and caps, non-compete scope. Market terms for a US middle-market deal under $50M.
12. Commercial Lease Red Flags
Review the pasted commercial lease from the tenant’s perspective. Flag: personal guaranty scope, CAM audit rights, relocation clauses, holdover penalties, assignment restrictions, and pass-through of landlord financing costs. Suggest three negotiation priorities.
13. IP Assignment Precision Check
Confirm whether the pasted IP assignment achieves a present-tense assignment (“hereby assigns”) versus an agreement to assign in the future. Explain the difference in one paragraph, then rewrite the clause to unambiguously effect a present assignment of all rights, including works made for hire and waivable moral rights.
14. Software Reseller Agreement Skeleton
Draft a reseller agreement framework: appointment scope (exclusive/non-exclusive, territory), pricing and margin, marketing obligations, end-customer flow-through, warranty pass-through, indemnity allocation between vendor and reseller, termination-for-cause triggers. Leave commercial numbers blank.
15. Boilerplate Cleanup Pass
Review the boilerplate section (governing law, venue, notices, entire agreement, severability, waiver, counterparts, e-signature). Flag any provision that conflicts with a more specific clause elsewhere. Suggest a clean, internally consistent boilerplate block.
12 Litigation Support Prompts
Litigation work has the highest hallucination risk because it involves case citations. Every prompt below assumes you will verify authorities against Westlaw, Lexis, or PACER before anything reaches a filing.
16. Discovery Response Draft — Requests for Production
I will paste a set of RFPs served on my client. Draft objections and responses for each. Use standard federal objections (overbroad, unduly burdensome, not proportional, vague, seeks privileged material, seeks trial-preparation material). Format: Request | Objection | Response. Do not cite specific cases — I will add authority.
17. Deposition Outline From Case Facts
Given the pasted case summary and witness role, generate a deposition outline for [witness]. Structure: (1) background & foundation, (2) admissions we need on record, (3) impeachment areas, (4) documents to authenticate, (5) closing loop-back questions. Aim for a 3-4 hour deposition.
18. Deposition Summary Reformat
I will paste a rough deposition summary. Reformat into: Topic | Page:Line | Key Testimony | Follow-up Needed. Preserve exact page:line citations. Flag any testimony that contradicts the witness’s prior statements listed in [prior statements].
19. Timeline From Fact Pattern
Build a chronological timeline from the pasted fact statement. Columns: Date | Event | Actor | Source Document. Flag any date where a source is not identified. Preserve exact dates — do not infer.
20. Meet & Confer Letter Draft
Draft a Rule 37 meet-and-confer letter identifying the following discovery deficiencies: [list]. Professional tone, specific citations to the deficient requests, request a call within 7 days, preserve position on motion to compel. No threats.
21. Motion in Limine Framework
Draft the argument framework (not authority) for a motion in limine to exclude [evidence type]. Structure: (1) evidence at issue, (2) proposed use by opposing party, (3) grounds for exclusion under FRE 401, 403, 404, or 702 as applicable, (4) prejudice analysis. Leave citation placeholders as [CITE].
22. Brief Section — Statement of Facts
From the pasted record, draft the Statement of Facts for a summary judgment brief for the [movant / non-movant]. Every factual sentence must include a citation placeholder like [Ex. A at 12]. Do not add facts not in the record. Neutral but favorable framing.
23. Opposition Brief — Counter-Argument Skeleton
I will paste the opposing party’s brief. Identify their three strongest arguments and their weakest. For each, draft a two-sentence counter-argument framework. Do not draft citations. Flag any factual claim in their brief that appears unsupported by the record I’ve pasted.
24. Client Litigation Status Update
Draft a monthly litigation status update for [case name]. Cover: what happened this month, what is scheduled next month, current budget-to-actual, any strategic decisions the client needs to make. Under 350 words. Business tone.
25. Expert Witness Preparation Q&A
Generate the 20 hardest cross-examination questions our expert should expect on [subject]. For each, note the concession opposing counsel is fishing for. Do not draft answers — that is the expert’s job.
26. Settlement Demand Letter Framework
Draft a settlement demand letter framework: liability summary, damages breakdown (economic, non-economic, punitive if available), settlement figure, response deadline, litigation posture if rejected. Firm but not aggressive. Leave damages numbers as placeholders.
27. Jury Instructions Cross-Check
Compare the pasted proposed jury instructions against pattern instructions for [jurisdiction, cause of action]. Flag any deviation from the pattern, any element misstated, and any instruction that shifts a burden of proof. Identify issues only — do not draft new instructions.
10 Client Communication Prompts
28. Intake Summary From Consultation Notes
From my pasted consultation notes, produce a structured intake summary: client goals, key facts, potential claims or defenses, jurisdictional issues, immediate deadlines, information still needed, conflict-check items. Neutral tone. No legal conclusions yet.
29. Engagement Letter First Draft
Draft an engagement letter for a [practice area] representation. Include: scope, fee structure ([hourly/flat/contingency]), retainer and replenishment, communication expectations, termination rights, file retention, arbitration or venue. State: [state]. Placeholders for numbers and parties.
30. Retainer Explanation for a Nervous Client
Rewrite the pasted retainer agreement into a plain-English one-page explainer for a small-business client who has never hired a lawyer. Factual. Do not oversell. Address the three fears every first-time client has: cost, control, and confidentiality.
31. Bad-News Letter
Draft a client letter delivering bad news: [summarize development]. Honest, direct, non-defensive. Structure: what happened, what it means, what we’re doing about it, what decision the client needs to make and by when. Under 300 words.
32. Statute or Regulation Translation
Translate the pasted statute/regulation into plain English for a business owner. Keep the structure of the original (elements, exceptions, penalties) but strip the legalese. If a term has specialized legal meaning that cannot be simplified, keep it and add a one-sentence gloss.
33. Court Ruling Explainer
Explain the pasted court ruling to my client in under 250 words. Cover: what the court decided, why, what it means for us, what happens next. No hedging language unless the ruling itself is ambiguous.
34. Fee Estimate Framework
From the pasted matter scope, generate a phased fee estimate: Phase | Activities | Estimated Hours | Estimated Fees | Assumptions. Include a paragraph on what could cause the estimate to increase. Assume [hourly rate] blended.
35. Meeting Recap Email
From the pasted meeting notes, draft a recap email to the client: decisions made, action items with owners and deadlines, open questions. Under 200 words. Bullet-heavy.
36. Client FAQ for a Practice Area
Generate 15 FAQs a prospective client in [practice area] would ask before hiring. For each, draft a 40-60 word answer in the voice of a trusted advisor, not a marketer. No guarantees. No superlatives.
37. Matter Closing Letter
Draft a matter-closing letter: summary of representation, final invoice reference, file retention terms, statute of limitations reminders for anything the client should track post-representation, thank-you close. Firm-formal tone.
8 Marketing & Business Development Prompts for Attorneys
Bar advertising rules apply. Every output below needs a compliance pass under your state’s Rules 7.1–7.5 equivalents before it goes live.
38. Practice Area Landing Page Draft
Draft a 600-word landing page for my [practice area] practice. Structure: headline, sub-hook, who we serve, what we do, how we work, why clients hire us, next-step CTA. No superlatives ("best," "top-rated") without a factual basis. No client outcomes without a disclaimer.
39. LinkedIn Article Outline
Give me a LinkedIn article outline on [legal development] for a [target industry] audience. Structure: hook, what changed, what it means practically, three concrete steps, close. Under 900 words when written. Voice: knowledgeable colleague, not thought-leader-on-stage.
40. CLE Presentation Framework
Build a 60-minute CLE presentation outline on [topic]. Include: learning objectives, section breakdown with timing, three case studies (leave as placeholders), audience Q&A prompts, one polling question. Assume intermediate audience.
41. Firm Newsletter Blurb
Write a 150-word newsletter blurb summarizing [recent case or regulation]. Business audience. End with a one-sentence "what to watch" and a soft CTA to contact the practice group.
42. Referral Follow-Up Email
Draft a follow-up email to a referral source who sent us a client that closed. Warm, specific, not gushing. Reference the matter type generically for confidentiality. Suggest a next touchpoint.
43. Speaking Pitch Email
Draft a pitch email to a conference organizer proposing a 45-minute session on [topic]. Include: proposed title, three learning objectives, why I’m qualified in two sentences, one prior speaking credential. Under 200 words.
44. Attorney Bio Rewrite
Rewrite the pasted attorney bio in third person, 200-250 words. Structure: what they do, who they do it for, one signature win phrased generically, credentials, one humanizing sentence. No "passionate," no "results-driven," no "trusted advisor."
45. Client Case Study (Anonymized)
From the pasted matter facts, draft an anonymized case study: the challenge, the approach, the outcome, the takeaway. Change identifying details. Include a bar-compliant disclaimer that prior results do not guarantee similar outcomes.
Best AI Tools for Lawyers in 2026
The 2026 tool landscape has consolidated around four categories: general-purpose assistants (ChatGPT, Claude), research tools (Perplexity), legal-specialized platforms tied to actual case databases (Lexis+ AI, Westlaw Precision AI, Thomson Reuters CoCounsel), and enterprise legal-AI platforms sold to BigLaw and in-house teams (Harvey and competitors). Positioning matters more than headline features.
| Tool | Best For | Citation Verification | Confidentiality Posture | Realistic Fit |
|---|---|---|---|---|
| ChatGPT Team / Enterprise | Drafting, summarization, client communication, marketing | None — treats citations as text | Team & Enterprise tiers do not train on your data by default | Solo through BigLaw for non-research tasks |
| Claude (Team/Enterprise) | Long-document analysis, contract review, nuanced drafting | None — same caveat as ChatGPT | Team & Enterprise tiers do not train on your data by default | Strong for contract-heavy transactional practices |
| Perplexity Pro | Legal-news research, secondary-source discovery | Web citations shown, not case-law grounded | Standard SaaS; not designed for privileged data | Background research only, never filings |
| Lexis+ AI / Westlaw Precision AI | Case research, jurisdictional analysis, brief drafting with authority | Yes — grounded in the vendor’s case database | Enterprise-grade, designed for law firm data | Any US practice doing meaningful case-law work |
| Thomson Reuters CoCounsel | Document review, deposition prep, contract analysis | Yes for research tasks tied to Westlaw | Enterprise-grade legal SaaS | Mid-size firms and up |
The honest read: ChatGPT and Claude are exceptional for the 70% of legal work that is drafting, translating, summarizing, and structuring. They are dangerous for the 30% that requires verified authority. A firm that pairs ChatGPT Team with Lexis+ AI or Westlaw Precision AI covers both jobs at a fraction of the cost of a single enterprise legal-AI seat.
Pro Tips, Trivia & Warnings
💡 Pro Tip #1: The Two-Model Verification Pattern
For any substantive output, run the same prompt through ChatGPT and Claude, then compare. Divergences flag hallucination risk faster than any single-model review. Ninety seconds, catches roughly half the fabricated facts and citations we see in real practice.
💡 Pro Tip #2: Redact Before You Prompt, Even on Enterprise Tiers
Enterprise tiers do not train on your data, but the log still exists on a third-party server. For the highest-sensitivity matters — anything under seal, matters involving minor clients, anything under a heightened protective order — use fictionalized facts even on Team/Enterprise. Cheap insurance.
🎓 Did You Know?
ABA Formal Opinion 512, issued July 2024, was the first comprehensive ABA guidance on generative AI in law practice. It addresses competence, confidentiality, communication with clients about AI use, candor to the tribunal, supervision, and fees. If you have not read it, that is your Rule 1.1 homework this quarter.
⚠️ Warning: The Mata v. Avianca Precedent
In June 2023, Judge P. Kevin Castel of the Southern District of New York sanctioned two attorneys for filing a brief containing six fabricated case citations produced by ChatGPT. They claimed they did not know the tool could hallucinate. That defense failed. Sanctions, public reprimand, mandatory CLE, and a permanent citation in every legal-AI ethics deck since. Verify every authority against a real database. Every time.
Honest Limitations
Hallucinated case citations. The number-one malpractice vector. General-purpose models produce citations that look correct — proper Bluebook form, plausible reporter numbers, real-sounding case names — that do not exist. State and federal dockets in 2024 and 2025 saw more than a dozen sanctions orders on the Mata v. Avianca fact pattern. Verify every citation against Westlaw, Lexis, or PACER.
Jurisdictional drift. Ask ChatGPT for the non-compete standard in North Dakota and you may get California law. The model does not always know what it does not know. Every jurisdiction-specific output needs a state-bar or reporter cross-check.
Privilege risks with third-party AI. Putting privileged content into a third-party system creates arguable waiver exposure in some jurisdictions, particularly if vendor terms allow any human review. Enterprise contracts with no-training and no-human-review provisions substantially reduce the risk but do not eliminate the argument.
Stale training data. Any general-purpose model has a knowledge cutoff. A statute amended six months ago, a circuit split resolved last quarter, an agency rule finalized yesterday — none of it is guaranteed to be in the training set. Real-time legal research belongs in a tool designed for it.
Client consent under evolving rules. ABA Formal Opinion 512 and several state bar opinions require lawyers to consider whether AI use on a matter requires client consent, particularly when client-identifying information is shared with the tool. Build that check into intake, not after the fact.
FAQs
Can I use ChatGPT for client work?
Yes, on a data-protected tier (ChatGPT Team, Enterprise, or the API), with client-identifying facts redacted, and with a verification plan for every claim in the output. Under ABA Formal Opinion 512 that combination satisfies competence and confidentiality — but verification is your responsibility.
What happens if ChatGPT hallucinates a case citation and I file it?
Sanctions. The Mata v. Avianca attorneys were fined, publicly reprimanded, and required to notify the judges falsely cited. Multiple later orders have imposed similar or steeper sanctions. Verify every cite.
Is prompting client information into ChatGPT a privilege waiver?
Possibly, and jurisdiction-dependent. Enterprise-grade tools with no-training and no-human-review terms substantially reduce the argument, but no appellate court has definitively resolved the question. Best practice: fictionalize facts, or use a legal-specialized platform with vendor-side privilege protections.
Which model is safest for legal drafting?
ChatGPT (Team/Enterprise) and Claude (Team/Enterprise) are roughly equivalent on quality. Claude’s larger context window is slightly better for long-document contract review. Neither is safe for unverified case research — for that, use Lexis+ AI, Westlaw Precision AI, or CoCounsel.
How do I meet ABA Model Rule 1.1 competence for AI?
Three steps: (1) read ABA Formal Opinion 512, (2) understand how the tools you use handle your data and where they fail, (3) build verification into your workflow. Take a generative-AI CLE annually and document your firm policy in writing.
Should my firm have a written AI policy?
Yes. A minimum-viable policy covers approved tools, prohibited use cases, data-handling rules, verification requirements, client-consent triggers, and a supervising-lawyer sign-off protocol. One page, reviewed annually.
Can I bill clients for time saved by AI?
Under Formal Opinion 512, hourly billing for time not actually spent is not permitted. If AI cut a task from six hours to ninety minutes, you bill ninety minutes. Flat-fee arrangements are a cleaner economic model for AI-augmented work.
Do I have to tell my client I’m using AI?
Often, yes. Formal Opinion 512 and several state opinions require disclosure when AI use is material to the representation or when client-identifying data is shared with the tool. When in doubt, disclose — discovery mid-matter is much worse than a proactive conversation.
What about opposing counsel using AI — can I use that against them?
Several federal district courts now require parties to certify whether filings contain AI-generated content. If you suspect fabricated citations in an opposing brief, verify each one and raise the issue with the court promptly. Judges have shown they take it seriously.
Where do I find prompt templates for my specific practice area?
Browse the free legal prompt library on PromptSpace and filter by task type. The 60 templates above are the starting set — the library adds specialized templates monthly across contracts, litigation, IP, employment, and regulatory practice.
Get the Full Prompt Library
Every prompt in this post is free to use, adapt, and share. For the full set — plus specialized templates for IP, employment, regulatory, and appellate work — browse the PromptSpace prompt library. New legal templates added monthly, all free, no email required.
Related Reading
References & Further Reading
- ABA Model Rules of Professional Conduct — Rules 1.1, 1.6, 3.3
- ABA Formal Opinion 512 on Generative AI (July 2024)
- Reuters: New York lawyers sanctioned for using fake ChatGPT cases — Mata v. Avianca, S.D.N.Y. 2023
- LexisNexis: Lexis+ AI Overview
- Thomson Reuters: CoCounsel
