Frequently Asked Questions

1. Why does definitions check on a legal document matter?

A typical 150+ page legal agreement may have about 25-40% of the pages devoted solely to definitions and may contain 400+ defined terms. A defined term that is misused or unused, or use of terms that are undefined creates ambiguities that attorneys are trained to avoid, not to mention the impression of sloppiness that detracts from its quality and user friendliness; it is considered best practices within the legal community. Moreover, defined-terms-related errors have lead to costly lawsuits.

2. What are the requirements for the definitions check app?

First, the document to be analyzed must be in the Microsoft Word format. The app is able to handle different versions of Word, including both DOC and DOCX formats, but is unable to analyze WordPerfect WPD, Google Docs, Adobe PDF or TXT files. Second, the document needs to have defined terms within quotation marks, either in single (') or double quotes ("); it is also able to handle smart/curly quotation marks (' ' " "). This is how the app is able to locate the defined terms.

3. What does the definitions check app do exactly?

The app highlights defined terms in legal documents to assist the reviewing attorney in identifying errors related to defined terms. First, it searches for defined terms, which are identified by being capitalized and enclosed in quotation marks. Then it highlights in Green all of the terms that are defined and used within the body of the document. Plural and singular forms of the defined term are also highlighted green if a match exists within the document. If no match is found, the quoted defined term is highlighted in Red to alert the reviewing attorney. When performing these checks, all lower cased quoted terms such as 'as is' are ignored.

Second, the app scans the document for Capitalized terms that are not defined in the document and are also not generally known as proper nouns within the legal community. Once a term has been detected, it is highlighted in Pink.

At the user's option, the app searches for lower-cased versions of defined terms within the document and highlights each of them Yellow if a match is found. This technique is useful when the reviewing attorney wishes to confirm that the lower casing of the terms was intentional.

When the app completes the highlighting algorithm, the app produces the highlighted Microsoft Word document, together with a findings report in the Results panel. The report lists every potentially undefined term and unused definition with page numbers, the surrounding context, how often the term appears (including lower-cased uses), and — where one exists — the defined term you probably meant (e.g., "Administration Questionnaire" flagged, with "Administrative Questionnaire" suggested). Items are ordered by likely importance, so the most actionable findings appear first.

4. How does DefCheck avoid drowning me in false positives?

This is where most of our engineering effort has gone. Tables of contents, section headings, run-in headings ("Section 2.10. Optional Prepayments. The Borrower may..."), signature blocks and defined-term index tables are all detected and filtered automatically — even in documents converted from EDGAR HTML filings that carry no Word formatting cues. Terms quoted from external statutes ("as defined in the Code") are recognized as such rather than flagged. The remaining findings are ranked, with near-misses of your document's own defined terms at the top. On our benchmark of SEC-filed agreements hand-reviewed by a securities attorney, this filtering removes roughly three quarters of the raw flags while still surfacing over 98% of the issues the human review found. Like spell-check, some judgment calls remain — which is exactly what the findings report is designed to help you (or your AI assistant) resolve quickly.

5. How accurate is it, really?

We measure rather than guess. DefCheck's engine is continuously benchmarked against a corpus of real SEC-filed credit agreements, merger agreements and indentures that a practicing securities attorney reviewed by hand, finding several hundred genuine drafting errors. DefCheck surfaces over 98% of those expert-confirmed issues. Its "did you mean" suggestions are informed by a research dataset of more than 30,000 individually reviewed flags across 7,200 SEC-filed contracts — so common slips like "Issuing Bank" vs. "Issuing Lender," "Hazardous Substances" vs. "Hazardous Materials," or "Event or Default" are recognized from thousands of prior human corrections, not just string similarity.

6. Can I use the results with an AI assistant like Claude or ChatGPT?

Yes — the findings report is designed for exactly that. It ends with review instructions, so you can paste the entire Results panel into your AI assistant of choice and receive back a concise, reviewer-style memo ("pg. 63: 'Administration Questionnaire' is not defined; recommend using 'Administrative Questionnaire'"). In our blind tests, an AI assistant working from DefCheck's report reproduced roughly nine out of ten items on the expert attorney's final list for the same document — in about two minutes, including catches the human review missed. As always, the output assists your professional judgment; it does not replace it.

And note what you would be sharing: the findings report, not your document. The report contains only the flagged terms and a brief snippet of surrounding text for each — isolated fragments that are largely meaningless out of context and do not reconstruct the agreement, so pasting it into an AI assistant is a far narrower disclosure than uploading the document itself. (Your document never leaves DefCheck's check process in the first place — see #10 below.) As with any tool, follow your organization's policies on the use of AI assistants.

7. Why do you highlight the terms instead of giving users the option to change the terms in real time?

We chose highlighting over other forms of communicating error-related information because it is consistent with how some attorneys already do it when performing the check manually, does not make any substantive changes to the document and is easy to reverse if necessary. An interface similar to "spell check" function of Microsoft Word that allows real-time interaction can be helpful, but in our view, has its own issues because it forces users to make the decision on the spot, when there are multiple ways to resolve a defined-term-related error, such as lower-casing the term, creating a new defined term, or using an already defined term. Best solution is possible only after having identified the errors from start to finish and the user is armed with information relating to the type, frequency and magnitude of the errors. Plus, this type of design permits features like the Paired Document Check, as described in #9 below.

8. How long does a typical check take?

It depends on the length of the agreement and how many defined terms it contains, but on shorter documents of less than 50 pages, it would be a few seconds, but for 200+ page documents with several hundred defined terms, it could take about 2-3 minutes. Reviewing the processed document by an attorney will take some additional time.

9. What is a Paired Document Check?

This feature will come in handy when you come across transactions where one document relies on defined terms contained in another document. A common example would be a collateral agreement in a loan transaction, which may contain a phrase to the effect of: "Capitalized terms used in this Collateral Agreement and not otherwise defined herein shall have the respective meanings ascribed to them in the Loan Agreement." Another example may be that of a Fee Letter, relying on the definitions contained in the Commitment Letter or of a Supplemental Indenture, relying on the definitions in the Base Indenture. In such cases, our Paired Document Check will eliminate the pain of having to flip back and forth between two documents - it will consider defined terms contained in the Source File when performing the definitions check on the Target File. In other words, Source File's defined terms (both singular and plural forms) will be highlighted green in the Target File and be considered when making other highlights.

10. How secure is your website?

Our app is designed by an experienced attorney who understands that documents come with both information and responsibility. Our service is hosted on Amazon Web Services (AWS), whose data centers have received ISO 27001, PCI DSS, SOC 1 Type II and SOC 2 Type II certifications. We utilize the industry standard SSL/TLS encryption so that the communication between the server and client browser remains private. We immediately and permanently delete the uploaded document upon completion of the check process - in fact, our serverless architecture processes your document in memory and never writes it to disk. Our goal is to eliminate any possibility of document retention since such procedure exposes us to legal risks. And we never share any of our customer information to third party vendors for marketing purposes - we hate spam just as much as you do!

11. Is this really free?

Yes! We have moved DefCheck.com onto a serverless architecture that allows us to provide this service free of charge. We hope you find it useful in your daily practice.

12. Do you have a corporate edition of this app? Will you be able to implement a feature I need? I have other questions!

If you have any suggestions or need further assistance, please contact us at legal@celant.com.