Pre-Filing Citation Check: How to Verify Legal Citations Westlaw and Lexis Don't Catch
Research platforms assume the citation exists. The author found the case, opened the opinion, and added the cite. The platform then renders the surrounding context, suggests Bluebook formatting, and links to the next reference. That workflow assumes the input.
A pre-filing citation check does not assume the input. It verifies the citation against reporter and public-record sources before the brief is filed. The output is an audit trail recorded as the verification ran, not a research result.
The distinction matters more in 2026 than it did in 2022, because more first drafts now arrive on a partner's desk after passing through a generative model. The audit layer is the step that closes the loop between drafting and filing.
The fundamental difference between legal research and the audit layer
Legal research is the activity of finding authority that supports a position. Westlaw, Lexis, and CourtListener are research tools. Their function is to surface relevant case law, statutes, and secondary sources. They assume the user is starting from a question and looking for an answer.
The audit layer is the activity of confirming that the authorities already cited in a draft can be located in their reported source. The function is different. Verification works backward from the cite to the reporter, not forward from the question to the cite. A research platform is not built for that direction of work, and was never meant to be.
When a firm runs a Filing Risk Scanner on a draft, the scanner is not asking what is the best case for this position. The scanner is asking whether the case named in the footnote appears in the reporter as cited, and whether the cited language matches the language in the reporter.
Those are two different jobs.
How standard research platforms assume citation existence
Open Westlaw to a citation page. The platform renders the opinion, the parallel cites, KeyCite history, and headnotes. The premise is that you are reading a case that exists.
Now consider the inverse workflow. A draft arrives with twelve footnoted citations. Three of them came from a generative model. Two of those three exist in the form cited. One does not appear in the reporter the cite names. Westlaw cannot tell you which one could not be located, because Westlaw assumes you found the case before you started.
Citation formatting tools have the same limit. A Bluebook formatter normalizes the cite to the correct typography. The formatter does not evaluate whether the underlying case can be located. The check happens at the layer below format.
The separation is structural, not a deficiency. Research tools are built to find. Audit tools are built to verify. A firm needs both, sequenced, and the audit step runs after drafting and before filing.
The risk profile of pre-filing workflows in AI-assisted drafting
The risk profile of a brief that included generative drafting is not catastrophic, and the discipline to manage it is not unusual. The risk is concentrated in one specific failure mode: a citation that could not be verified in its reporter, included in a filed brief, surfaced later by opposing counsel or the court.
Bar opinions issued across 2024 and 2025 have repeatedly framed the duty as a duty to verify, not a duty to abstain. Standing orders requiring AI-use disclosure have appeared in multiple federal districts. The institutional posture is converging on the view that AI use is permitted, and citation verification is the affirmative obligation.
A pre-filing citation check is the operational answer to that obligation. The firm has used the tool that helped it draft. The firm has run the audit that confirmed the citations. The matter file contains the record of both.
Moving from manual partner review to a scalable Filing Risk Scanner
The traditional answer to citation risk is partner review under Rule 11. The partner reads the brief, recognizes the cases, and notices when a cite reads off. The method works when the partner has read every cited opinion in the last five years. The method scales poorly when the volume of AI-assisted drafts increases.
A Filing Risk Scanner does not replace the partner. The scanner extracts the citations from the draft, resolves each one against reporter and public-record sources, classifies each by verification status, and records the result. The partner then reviews the scanner output, which is a structured list rather than an unstructured brief.
The verbs change. The partner moves from finding the problems to confirming the resolutions. That is a more scalable job. The scanner does not exercise legal judgment. The scanner confirms or hedges on existence. The partner retains every judgment call.
Generating a hashed, time-stamped Verification Certificate prior to filing
The final step in the pre-filing check is the generation of the artifact. The Verification Certificate contains the citations that were checked, the verification status for each, the source records consulted, and a SHA-256 hash of the certificate content. The certificate carries a public token URL and a timestamp.
The certificate exists for three audiences. The first is the partner signing the brief, who reads it before signing. The second is the firm's matter file, where the certificate is filed alongside the brief and the engagement letter. The third is the future reader who may, in a malpractice review or a sanctions hearing, need to confirm that citation verification was performed and what it found.
A hashed, time-stamped certificate is the difference between a verification that happened and a verification that can be proven. Both are useful. Only one is auditable.
The Filing Risk Scanner is available at veritaslaw.app. The certificate is generated per filing. The scanner does not write the brief, format the citations, or argue the position. The scanner verifies the cites and produces the record.