Contract Review Prompts for Lawyers: 20 AI Prompts for Issue Spotting

Lawyers get usable contract review from an AI model when the prompt names a role, gives the deal context, states one review task, and fixes the output format. Twenty contract review prompts follow, grouped by job: a master prompt for a full review of one agreement, four ChatGPT prompts for everyday review tasks, and a checklist prompt. Ten step-by-step prompts cover each clause group, a redline prompt compares two drafts, three advanced issue spotting prompts go deeper on risk, and the steps to validate any AI edit round out the set before anyone relies on it. Generic instructions such as “review this contract” return shallow summaries because the model has no role, no context, and no output specification.

The prompts below are written for commercial lawyers working through third-party paper, and each one carries bracketed fields for the contract type, the side you represent, and the governing law. None of them names a model-specific feature, so they are model-agnostic, and they are the manual equivalent of what the Review feature does on a first pass.

Key takeaways

Contract review prompts return a usable result when they name a role, state the deal context, assign one task, and fix the output format, rather than asking a model to review a contract. Every AI finding needs a section number a lawyer can check against the executed text.

1. A redline prompt compares two drafts for substantive changes only, filtering out formatting and renumbering noise.
2. A four-step validation pass checks every AI finding against the executed text before anyone relies on it.
3. AI assists by producing actionable, business-focused findings, not by replacing a lawyer’s judgment on what those findings mean for the deal.

[cta-1]

The master contract review prompt

A master contract review prompt replaces four or five separate instructions with one block: role, deal context, review tasks, output format, and the contract text. Fill the bracketed fields before pasting, and the model returns an executive summary, a prioritized risk list, and proposed redlines in one pass.

```

ROLE: Act as a commercial contract lawyer representing [THE SIDE YOU REPRESENT] in a [CONTRACT TYPE].

CONTEXT: The counterparty has [COUNTERPARTY LEVERAGE]. This agreement is governed by [GOVERNING LAW], and the outcome we need is [COMMERCIAL OUTCOME].

TASK: Review the contract text below and return:

1. An executive summary of the parties, purpose, term, and key financial obligations.
2. A risk and exposure analysis, with each finding rated High, Medium, or Low.
3. Proposed redlines, with the section number and one “Give” and one “Get” for each.

OUTPUT FORMAT: A client-ready memo with the three numbered sections above.

CONTRACT TEXT: [PASTE THE AGREEMENT]

```

Five fields drive the result:

1. Contract type: name the agreement, such as an MSA (master service agreement), an NDA (non-disclosure agreement), a vendor agreement, or a lease, so the model does not default to generic contract norms.
2. The side you represent: changes which provisions get flagged as risk.
3. Counterparty leverage: a term that is fine from a position of strength is a risk from a position of weakness.
4. Governing law: the jurisdiction behind every enforceability judgment the model makes.
5. The commercial outcome you need: what the deal is actually for, not just “a good contract.”

The same five-part shape, with the task line swapped for the legal task at hand, is the general ChatGPT prompt template for lawyers, and it works on any legal task, not just contract review.

ChatGPT prompts for contract review

Four ChatGPT prompts for contract review cover the tasks a lawyer runs on a first pass over third-party paper: an executive overview, a risk and exposure analysis, a missing-clause and gap check, and a proofreading and quality check for defined terms and cross-references.

Prompt 1: Executive contract overview

```

Analyze this contract and provide an executive summary covering: (1) parties and their roles, (2) contract purpose and scope, (3) term and termination rights, (4) key financial obligations, (5) major liability provisions. Format as a client-ready executive summary.

```

Output spec:

1. Parties and their roles
2. Contract purpose and scope
3. Term and termination rights
4. Key financial obligations
5. Major liability provisions

Prompt 2: Risk and exposure analysis

```

Review this contract for provisions that create risk for [CLIENT X]. Identify: (1) unlimited or uncapped liability, (2) broad indemnification provisions, (3) unfavorable termination or renewal terms, (4) ambiguous or one-sided language, (5) missing limitation of liability protections. Rate each finding High, Medium, or Low, and prioritize findings by potential financial impact.

```

Output spec:

1. Unlimited or uncapped liability, rated High, Medium, or Low
2. Broad indemnification provisions, rated High, Medium, or Low
3. Unfavorable termination or renewal terms, rated High, Medium, or Low
4. Ambiguous or one-sided language, rated High, Medium, or Low
5. Missing limitation of liability protections, rated High, Medium, or Low

Prompt 3: Missing clauses and gap check

```

Compare this vendor agreement against the clauses a vendor agreement of this type normally carries: indemnification, insurance requirements, data security obligations, service levels, and termination for the vendor’s non-performance. List every clause that is missing or addressed only partially, and explain what gap each omission creates.

```

Output spec:

1. Clause expected but missing, with the gap it creates
2. Clause present but only partially addressed, with what is missing from it
3. Recommended addition, in one sentence

Prompt 4: Proofreading and quality check

```

Proofread this contract for drafting errors: capitalized terms used but never defined, terms defined but never used, cross-references to sections that do not exist or were renumbered, and leftover template language such as bracketed placeholders or a prior party’s name. Cite the exact section for every issue found.

```

Output spec:

1. Undefined capitalized terms, with the section where each appears
2. Defined-but-unused terms, with the section where each is defined
3. Broken or missing cross-references, with the section that cites them
4. Leftover template content, with the section where it appears

Spellbook’s analysis of 3,019 material-contract exhibits filed with the Securities and Exchange Commission (SEC) on EDGAR found that 60% of SEC-filed contracts contain drafting errors like these, which is the scale of the problem a dedicated proofreading pass catches that a single read-through does not. Spellbook Labs has the full study.

Contract review checklist prompt for systematic analysis

Using a contract review checklist can help keep every category in scope. This prompt produces a structured report that covers all major contract elements.

Complete contract review checklist prompt

```

Perform a thorough contract review using this checklist with findings for each category:

1. Parties and effective date: verify proper legal entities and execution.
2. Term and renewal: assess length, renewal mechanisms, and termination rights.
3. Payment terms: extract pricing, payment schedule, and adjustments.
4. Obligations: identify each party’s deliverables and performance standards.
5. Warranties: review representations, warranties, and disclaimers.
6. Indemnification: analyze indemnity scope, procedures, and limitation of liability.
7. IP and confidentiality: evaluate IP ownership and licensing terms and confidentiality scope.
8. Termination: analyze termination and exit provisions for cause, convenience, and breach.
9. Dispute resolution: identify jurisdiction and governing law issues.
10. General provisions: review assignment, amendment, notices, compliance.

For each category, flag issues requiring attention and provide recommendations.

```

Run this contract review checklist against the redlined draft at each round of negotiation, not only the signature-ready version. A category marked clear on the first pass can need a second look after the other side proposes changes to payment terms or limitation of liability, since those edits often ripple into clauses the checklist already closed out.

Step-by-step contract review prompts

Breaking the contract review into steps allows a lawyer to analyze each element more thoroughly, with more extensive results than a single prompt returns. For Step 6 through Step 10, start the prompt with: “Building on the previous analysis of [Party A] and [Party B], evaluate...” so the model carries the earlier analysis forward.

Step 1: Parties, recitals, and definitions review

```

Examine the following contract sections: (1) identify all parties, their legal status, and authority to contract, (2) review recitals for purpose and context, (3) extract all defined terms and assess clarity, (4) flag ambiguous definitions or missing key terms, (5) flag capitalized terms that are not defined, (6) spot potential interpretation issues.

```

Step 2: Term, termination, and renewal analysis

```

Analyze termination and exit provisions in this contract: (1) contract term and commencement date, (2) termination rights for each party (cause vs. convenience), (3) notice requirements, (4) automatic renewal provisions, (5) post-termination obligations, (6) survival clauses. Track critical dates and renewal terms and flag unfavorable provisions.

```

An automatic renewal notice window that has already lapsed by the time anyone reviews the contract is a common and costly miss: the prompt should flag the date the notice window closes, not just that one exists.

Step 3: Payment terms and financial obligations

```

Extract and analyze all financial clauses: (1) pricing structure and amounts, (2) payment schedule and milestones, (3) late payment penalties and interest, (4) expense reimbursement, (5) fee adjustment mechanisms, (6) financial caps or limits. Assess financial obligations and liabilities and highlight risk factors.

```

Step 4: Performance obligations and deliverables

```

Review performance provisions: (1) scope of work or deliverables, (2) performance standards and metrics, (3) timelines and deadlines, (4) acceptance criteria, (5) change order procedures, (6) performance warranties, (7) consequences of non-performance. Identify unclear or unenforceable terms.

```

Step 5: Representations, warranties, and disclaimers

```

Analyze representations and warranties: (1) representations made by each party, (2) express and implied warranties, (3) disclaimer language and scope, (4) warranty duration and remedies, (5) “as is” provisions, (6) compliance representations. Flag one-sided or unusually broad disclaimers.

```

A disclaimer that reads as routine boilerplate can gut an express warranty made earlier in the same agreement, so the prompt should check the disclaimer language against every warranty it might cut back, not read it in isolation.

Step 6: Indemnification and liability provisions

```

Evaluate indemnification and liability: (1) indemnification obligations for each party, (2) scope of indemnity coverage, (3) indemnification procedures and notice, (4) limitation of liability clauses, (5) liability caps and exclusions, (6) consequential damages waivers, (7) insurance requirements. Detect unfavorable indemnification language and produce comprehensive risk matrices. Compare the indemnity section against the limitation of liability section.

```

Broadening the indemnity obligation without adjusting the limitation of liability can leave a party exposed well past what either clause reads as providing on its own. Indemnity, as a legal concept, is the obligation to compensate another party for specified losses, which is the baseline the prompt’s output should be checked against.

Step 7: Intellectual property rights

```

Evaluate IP ownership and licensing terms: (1) ownership of deliverables and work product, (2) IP licenses granted, (3) license scope and limitations, (4) IP warranties and indemnification, (5) confidentiality obligations, (6) restrictions on use, (7) IP developed during the term. Identify unclear ownership or excessively broad licenses.

```

Step 8: Confidentiality and data protection

```

Review data privacy and security obligations: (1) confidentiality definitions and scope, (2) obligations of the receiving party, (3) exclusions from confidentiality, (4) duration of obligations, (5) data protection and privacy requirements, (6) permitted disclosures, (7) return or destruction of information. Flag gaps in protection.

```

Step 9: Dispute resolution and governing law

```

Identify jurisdiction and governing law issues: (1) applicable law and jurisdiction, (2) dispute resolution method (litigation, arbitration, mediation), (3) arbitration provisions and rules, (4) venue selection, (5) waiver of jury trial, (6) attorney’s fees provisions, (7) pre-litigation requirements. Assess favorability to the client, and check for equitable relief provisions.

```

A venue clause that conflicts with an arbitration clause elsewhere in the same agreement is worth flagging on its own, since it signals the agreement was assembled from more than one template and other sections may carry the same mismatch.

Step 10: General provisions and compliance

```

Analyze general provisions: (1) assignment and change of control, (2) amendment procedures, (3) notice provisions, (4) force majeure, (5) severability, (6) entire agreement clause, (7) compliance with laws, (8) export control, (9) sanctions, (10) anti-corruption provisions. Flag missing protections.

```

How to compare two contract versions with a redline prompt

An AI prompt to compare two contract versions asks the model for substantive changes only, not formatting differences. Give it both drafts, name the side you represent, and require a table listing every changed provision by section number, whether the change favors your side, and any new risk it introduces.

```

Compare Draft A and Draft B of this agreement. Identify every substantive change: a provision added, removed, or reworded in a way that changes rights, obligations, or risk. Ignore formatting, renumbering, and non-substantive wording changes. For each change, state whether it favors [THE SIDE YOU REPRESENT] or the counterparty, and flag any new risk it introduces.

```

Output spec:

1. Changed provision and its section number
2. Favorable, unfavorable, or neutral to the side represented
3. New risk the change introduces, if any
4. Suggested response to the change

A model asked to compare two drafts of the same agreement tends to report renumbering and formatting differences as substantive changes unless the prompt rules them out explicitly, which buries the real changes in noise a lawyer then has to filter out by hand. Run the prompt at every exchange of drafts, not only on the final markup: a term introduced in one round and quietly reverted in the next is easy to miss without a side-by-side comparison at each step.

Which version’s terms control when the two drafts conflict is a legal question the comparison output does not answer. Under UCC § 2-207, a last-minute change exchanged between the parties can fail to become part of the contract at all, independent of what the model flags as a risk.

Advanced issue spotting prompts

These AI issue spotting prompts for contracts go past clause-by-clause review to the three questions a lawyer asks before a negotiation round: where the agreement is one-sided, where its own sections conflict, and where it fails a regulatory requirement. Each prompt names the output the model must return.

Negotiation leverage and position analysis

```

Assess this contract’s balance: (1) identify provisions favorable to each party, (2) evaluate overall fairness, (3) highlight significantly one-sided terms, (4) prioritize negotiation points by importance and likelihood of success, (5) suggest alternative language for unfavorable terms. Provide a negotiation strategy memo. For every unfavorable term identified, suggest one “Give” (a point that can be conceded) and one “Get” (what to demand in exchange).

```

Naming who the parties are sharpens this analysis: a ten-person startup facing a Fortune 500 counterparty has a different negotiating position than two parties of similar size, and the prompt should say so. A negotiation position holds better on proven alternative language from contract term precedents than on the model’s own suggested wording.

Cross-reference and consistency check

```

Identify conflicting or ambiguous terms: (1) internal inconsistencies between sections, (2) conflicting provisions, (3) undefined cross-references, (4) inconsistent terminology, (5) ambiguous language that could support multiple interpretations. Create an anomaly report with specific contract locations.

```

Compliance and regulatory gap assessment

```

Analyze the contract for regulatory compliance issues: (1) identify provisions that may violate applicable laws, (2) assess industry-specific regulatory requirements, (3) flag data privacy obligation concerns (GDPR, CCPA, or similar frameworks), (4) evaluate export control and sanctions compliance, (5) review licensing and permit requirements. Provide a detailed compliance risk assessment.

```

For industry-specific requirements such as HIPAA in healthcare or FINRA in finance, name the regulation specifically: a general compliance prompt can miss a nuanced state-level regulatory update.

A model asked whether a cap is market-standard answers from its training data, which is why clause benchmarking against a named contract set carries more weight than the model’s own impression.

How legal teams validate AI suggestions and edits to a contract

Legal teams validate AI suggestions and edits to a contract by requiring a section number with every finding, then checking each cited section against the executed text before accepting the edit. The validation pass has four steps: confirm the section exists and says what the model claims, confirm the defined terms the edit relies on are defined in the agreement, confirm the proposed language does not conflict with another provision, and record who approved the change.

1. Confirm the section exists, by opening the executed document at the exact section number the model cited and checking that it says what the model claims.
2. Confirm the defined terms are defined, rather than assumed from common usage: a term the model treats as understood may not appear in the agreement’s definitions at all.
3. Confirm no conflict with another provision, since a proposed edit to one section can contradict a term set elsewhere in the same agreement.
4. Record who approved the change, noting the section, the language accepted, who reviewed it, and the date, in whatever log or comment thread the review already uses.

That record is what lets a documented review process survive being audited later.

Append this line to any prompt on this page: cite the section number and quote the sentence you relied on for every finding. Skipping this pass is most tempting on findings that sound confident, which is exactly when a cited section that does not exist is easiest to miss.

The duty to verify AI output before relying on it is the same duty that governs any other research tool, and ABA Formal Opinion 512 names it directly for generative AI.

Best practices for using AI contract review prompts

AI contract review prompts work best alongside a lawyer’s own read of the agreement. The practices below hold across contract types:

1. Verify AI findings independently.
2. Use multiple prompts for a detailed review.
3. Customize prompts by jurisdiction and specific deal context.
4. Apply professional judgment to all flagged issues.
5. Anonymize client data before it ever reaches a prompt.

Anonymizing means removing client-identifying detail before pasting anything into ChatGPT or a similar consumer chat tool, which may retain what it is given.

Multiple prompts matter because each one targets a different failure mode: the executive overview will not surface a liability cap the risk-analysis prompt catches, and neither will flag a missing clause the gap-check prompt is built for. A prompt that is dialed in for a given contract type is worth the same care: it becomes one of the reusable legal prompts the team runs on similar agreements rather than rebuilding it from scratch.

Common mistakes to avoid in AI-assisted contract review

Filling in the master prompt’s role and side-represented fields but leaving the governing-law or commercial-outcome field blank is a common shortcut. The model then answers from generic contract norms instead of the specific deal, which defeats the reason to fill in any of the five fields at all.

Carrying the “Building on the previous analysis of [Party A] and [Party B], evaluate...” instruction from the step-by-step prompts into a fresh chat session is another mistake. With no earlier analysis in that session for the model to build on, it invents one rather than flagging that nothing came before.

Pasting an agreement long enough that the back half falls outside the model’s context window is a mechanical version of the same problem: sections the model never actually read can come back marked clear, which looks identical to a section that was checked and passed.

Running any of these prompts against a scanned or image-only PDF is a mechanical failure rather than a judgment error: the model reads whatever its text layer contains, which can be empty, garbled, or missing entire pages, not the actual agreement.

[cta-2]

How Spellbook handles contract review without prompt engineering

Spellbook handles contract review without any of the prompts above. Spellbook’s Review feature surfaces risks and proposes redlines as native Word tracked changes on every agreement, with no role, context, or task line to fill in first. The Playbooks feature applies a team’s own standards to every agreement automatically, and the Proofread feature catches the defined-term and cross-reference errors Prompt 4 above looks for by hand. The Compare to Market feature benchmarks a contract’s terms against Spellbook’s set of 20M+ real contracts, a different kind of market check than asking a general model what is standard.

Spellbook is a commercial-contract specialist, not a legal-research or litigation platform, so research memos and case-law analysis still go through other tools.

The test for any AI-assisted contract review tool is the same one that applies to the prompts above: every finding traces back to a section a lawyer can open and check.

Try Spellbook free.

Frequently asked questions

Can AI prompts replace manual contract review?

No. AI assists but does not replace lawyer review. Transactional lawyers must still assess legal meaning and make decisions based on client needs and relevant business settings. The prompts above work because they get the fundamentals right that any legal use of ChatGPT depends on: a role, a context, one task, and an output format.

Can ChatGPT be used for contract review?

Yes, within limits. ChatGPT can summarize a contract, flag obvious risk language, and answer questions about a pasted clause, as can Claude, Gemini, and Copilot. It does not replace a lawyer’s judgment on whether a clause is enforceable or acceptable for a specific deal, and output should be verified against the actual contract text before anyone relies on it.

Which contracts work best with AI review prompts?

Standard agreements such as NDAs, vendor contracts, employment agreements, and service contracts work best with AI review prompts, because the reviewer already knows the standard shape and can spot a deviation quickly. Heavily bespoke or heavily amended agreements work less well, since much of what matters sits in deal history outside the document itself.

What is contract risk analysis?

Contract risk analysis is the process of identifying provisions in an agreement that create financial, legal, or operational exposure, such as uncapped liability, broad indemnification, or unfavorable termination rights, and rating each one by how likely it is to matter and how much it could cost if it does.

Do I need different prompts for different contract types?

Yes. The role and task lines carry over between contract types, but the context line and the clause list have to be rewritten for each one, since what counts as a risk in an NDA is not what counts as a risk in a master service agreement.