To use AI for contract review safely: use an approved tool on terms that do not train on your data; remove identifying details where possible; give the AI your client’s role, the governing law and your playbook; ask for an issues list with clause references; ask for proposed wording separately; check every point against the actual contract text; and apply your own judgment on commercial context, enforceability and what is missing. AI speeds up the first pass, but the lawyer remains responsible for the advice.
What AI is good at, and where it slips
Good at: reading long agreements quickly, producing structured issues lists, comparing versions, summarising obligations and dates, suggesting alternative wording, and checking defined terms for consistency.
Where it slips: cross-references and definitions that change meaning elsewhere in the document, schedules it did not see, deal context it was never told, jurisdiction-specific enforceability, and the most important category of all: what is missing.
Before you start: two checks
- Is the tool approved for client documents? Use a business or enterprise plan or a legal tool whose terms prohibit training on your data. ABA Formal Opinion 512 indicates informed consent is needed before inputting confidential information into tools that may learn from it.
- Can you reduce what you share? Replace party names, prices and identifying details with placeholders using our Confidentiality Redactor, then restore them in the output.
The seven-step workflow
1. Brief the AI like a junior associate
Act as a senior commercial lawyer acting for the [customer] under [governing law].
Context: [one paragraph on the deal and the client's priorities].
Our positions: [e.g. liability cap at 12 months' fees; mutual indemnities;
30-day termination for convenience].
2. Ask for a prioritised issues list
Review the agreement below and produce an issues list grouped as High, Medium
and Low priority. For each: clause number, the issue, the risk to our client,
and whether it departs from our positions. Quote the clause text you rely on.
Do not propose wording yet.
3. Check every point against the text
Open each clause the AI cites. Confirm it quoted accurately and understood the definitions and cross-references. This is where most AI errors in contract work are caught.
4. Ask what is missing
List provisions commonly found in a [type of agreement] that are absent or thin
here, and explain why each might matter to our client.
Then run your own checklist. Absence is hard for AI and easy to miss under time pressure.
5. Decide what to raise
This is judgment, not automation: commercial leverage, relationship, risk appetite and market practice for this client.
6. Draft revisions
For issues 1, 3 and 5, propose revised wording that achieves our position,
in the drafting style of the agreement. Show original and proposed text and
explain each change in one sentence.
7. Final review
Read the marked-up agreement as a whole. Check numbering, defined terms, consistency between clauses and that your changes do not create new problems.
Five prompts worth saving
- Obligations tracker: “Extract every obligation, deadline and notice requirement into a table: party, obligation, trigger or date, clause.”
- Plain-English summary: “Summarise this agreement for a non-lawyer client in under 300 words: what they must do, what they get, key dates, how it ends, biggest risks.”
- Version compare: “List every change between these versions and its legal and commercial effect for our client.”
- Definitions check: “List defined terms used but not defined, defined but unused, or used inconsistently.”
- Negotiation playbook: “For each key clause, give our preferred position, fallback and walk-away point.”
More in the Prompt Library.
Choosing a contract AI tool
Word-based assistants suit lawyers who live in Microsoft Word; playbook tools suit teams with standard positions; lifecycle platforms suit in-house teams managing volume. Browse options in the legal AI tools directory and run every vendor through the due diligence questionnaire.
The bottom line
Let AI do the first read and the drafting legwork. Keep the decisions, the context and the final read for yourself. That is what clients pay for, and what professional rules require.
Frequently asked questions
Can AI replace a lawyer for contract review?
No. AI can produce a fast first pass, flag common issues and draft alternative wording, but it can miss context, misread cross-references, and lacks knowledge of the deal and client priorities. A lawyer must review and own the advice.
Is it safe to upload contracts to ChatGPT or Claude?
Only on an approved business or enterprise plan whose terms prohibit training on your data, consistent with your firm’s policy and client instructions. Consider redacting names and commercial terms, and obtain informed consent where required.
Which AI tools are built for contract review?
Options include Word-based assistants such as Spellbook, playbook-driven review tools such as LegalOn and Luminance, and contract lifecycle platforms such as Ironclad and Juro. General assistants on business plans are also widely used. See our tools directory.
Claude for Lawyers 2026
A focused guide to using Claude in legal work, from careful drafting and document review to client communication and repeatable workflows.
See it on AmazonTurn reading into a skill
Take the free 7-lesson course and earn a shareable certificate.
This article is general educational information, not legal advice. Rules differ by jurisdiction and change often; check the primary sources and your own professional rules before relying on it.