Practical product guide
AI contract review: what it finds, how it works, and where it stops
AI contract review turns an uploaded PDF or DOCX into a structured first-pass report: a 0–100 risk score, executive summary, key risks, clause findings, payment, termination and liability notes, missing protections, and recommended negotiation steps. It helps you triage a document and prepare questions, but it does not replace verification of the source contract or advice from a qualified lawyer.
Last reviewed:
The short answer
AI review is most useful before a human makes a decision. It can turn a long contract into a consistent checklist, surface language that deserves attention, and show which expected protections appear to be absent. The valuable output is not the score by itself; it is the traceable set of issues you can confirm, reject, or escalate. Treat every result as a hypothesis grounded in the text supplied to the model.
Contract Analyzer organizes the result into nine fields: risk score, summary, key risks, clauses, payment terms, termination terms, liability analysis, missing clauses, and recommendations. Clause findings receive low, medium, or high severity labels. Those labels prioritize reading; they do not measure enforceability, predict a dispute, or tell you whether a deal is acceptable for your organization.
How AI contract review works
The workflow has three visible stages. Each stage has a boundary that matters when you interpret the result.
- 1
Upload a supported contract
Upload a text-based PDF or DOCX file of no more than 10 MB. Encrypted PDFs are not supported.
- 2
Extract and analyze the text
Contract Analyzer extracts the document text and sends up to 12,000 characters to the configured AI provider for analysis.
- 3
Review the structured report
Read the risk score, summary, key risks, clause findings, missing protections, and recommendations while checking each important point against the original contract.
After the report is generated, review it beside the source document. Confirm names, dates, amounts, notice periods, renewal mechanics, defined terms, clause references, and exceptions. A concise report can make review easier, but it cannot restore text that was not extracted or included in the analysis context.
What the risk assessment means
The report assigns an overall integer score from 0 to 100, with a higher number indicating more detected risk. Read it as a prioritization device. Two contracts with the same score can create very different exposures, and a low score can still hide a deal-breaking term for your circumstances. The summary and key-risk list explain what drove the review and should receive more attention than the number alone.
A useful validation pass asks four questions: Is the cited language actually in the contract? Did the report account for definitions and cross-references? Does the issue affect your side of the deal? Is the suggested action commercially and legally appropriate in the relevant jurisdiction? If any answer is unclear, flag the point for a knowledgeable reviewer instead of treating the generated conclusion as settled.
Clause analysis and report fields
The clause list pairs a title with a finding and a low, medium, or high severity. The analysis is instructed to follow the document order where possible and to anchor important findings with a short quotation or exact section heading. Separate report fields focus on payment, termination, and liability because these terms often require a connected reading across definitions, exceptions, schedules, and surviving obligations.
Terms it summarizes
Price and currency, billing and due dates, late fees, refunds, taxes, termination rights, notice and cure periods, survival, liability caps, exclusions, indemnities, and carve-outs can appear in the dedicated fields when present in the supplied text.
Gaps it can flag
The missing-clauses list identifies protections that are absent or unclear, while the recommendations list turns detected issues into proposed questions, negotiation points, or redline directions. Absence still requires confirmation against the full document and every incorporated schedule.
Supported documents and input boundaries
The current upload flow supports PDF and DOCX files up to 10 MB. A PDF needs a usable text layer; encrypted PDFs are not supported. Image-only scans, damaged files, complex tables, headers, footnotes, tracked changes, or unusual layouts may not extract in the way a human sees them. Convert or clean the document before relying on an automated review if the extracted text looks incomplete.
Contract Analyzer stores and analyzes no more than the first 12,000 characters of extracted text in the current implementation. That means a long agreement may be truncated before later sections, exhibits, or signature blocks. Check the extracted-text view and split a long document into coherent parts when necessary, while preserving definitions and cross-references needed to interpret each part. See the current processing and retention details on the security page.
Limitations and responsible use
Contract Analyzer does not publish a validated accuracy rate. Accuracy varies with text extraction quality, document length, formatting, definitions, cross-references, and the context supplied to the configured model. Treat every finding as a review hypothesis, verify it against the source, and do not infer reliability from the risk score alone.
Generated analysis can omit terms, misread relationships, invent details, or apply a generic risk assumption that does not fit your deal. It does not know undisclosed facts, negotiation history, side letters, local law, regulatory duties, your risk tolerance, or whether another document changes the clause. Never copy a recommendation into a contract without checking how it interacts with the rest of the agreement.
The NIST AI Risk Management Framework is a voluntary framework for incorporating trustworthiness considerations into the design, use, and evaluation of AI systems. Its risk-management perspective supports a practical rule here: define the decision, understand the tool's limits, check the output, and keep a human accountable for the outcome.
For legal practice, ABA Formal Opinion 512 (PDF) explains that lawyers using generative AI must consider obligations including competence, confidentiality, client communication, supervision, candor, and reasonable fees. The opinion also emphasizes understanding a tool's capabilities and limitations. It is professional-responsibility guidance, not a product endorsement or a substitute for rules that apply in your jurisdiction.
AI contract review vs. lawyer review
| Question | AI first pass | Qualified lawyer |
|---|---|---|
| Organize the document | Summarizes supplied text into consistent report fields. | Connects text with schedules, facts, and the transaction. |
| Identify risk | Flags patterns and creates issues for verification. | Assesses legal effect, priorities, and jurisdiction-specific exposure. |
| Recommend action | Suggests questions and possible negotiation directions. | Advises, drafts, negotiates, and accepts professional responsibility. |
Use AI when you need a structured first look or a checklist for discussion. Escalate when the contract controls material money, intellectual property, personal data, employment, regulatory obligations, uncapped liability, a dispute, or rights you cannot easily undo. The strongest workflow combines machine-assisted organization with accountable human judgment.
AI contract review FAQ
What does an AI contract review include?
Contract Analyzer returns a 0-100 risk score, an executive summary, key risks, clause-level findings with low, medium, or high severity, payment terms, termination terms, a liability analysis, missing clauses, and recommended next steps.
Can AI contract review replace a lawyer?
No. AI contract review is a first-pass issue-spotting and organization tool. It cannot establish your legal position, understand every business fact, negotiate for you, or give advice tailored to a jurisdiction. Use a qualified lawyer for high-stakes decisions and verify important findings against the signed document.
Which contract files are supported?
The current upload flow accepts text-based PDF and DOCX files up to 10 MB. Encrypted PDFs are not supported. Scans without a usable text layer, unusual formatting, and content beyond the 12,000-character analysis limit may produce incomplete context.
How should I use the risk score?
Use the 0-100 score as a triage signal, not as a probability of loss, legal conclusion, or guarantee. Read the reasons beneath the score, compare them with the contract text, and decide which issues need negotiation or professional review.
More product, billing, and workflow answers are available in the full FAQ.
Start with a first-pass review, then verify
Compare plans before you upload, understand how contract data is handled, and keep legal review in the loop whenever a finding can materially affect your rights or obligations.