AI Disclosure for Legal Research and Law Review Submissions

Legal scholarship has a footnote problem. AI makes it sharper.

Legal writing asks readers to trust chains of authority.

A claim about doctrine points to a case. A claim about legislative intent points to a report. A claim about empirical legal studies points to data, code, interviews, or court records. When AI enters that chain, the reader needs to know where it helped and where the author checked the work.

That does not mean every spellcheck deserves a paragraph. It means legal scholars should record AI use when it touches research, citations, argument structure, drafting, translation, or revision.

If you used an AI tool while preparing a legal article, student note, book chapter, empirical legal study, or law review submission, create a record before you submit. The easiest way is to generate an [[AI Usage Card](/ai-disclosure-for-social-science-research/)](/) and keep it with your manuscript files.

Why legal research needs its own disclosure habit

Legal scholarship differs from many academic fields because authority sits inside the prose.

A footnote can carry the weight of a paragraph. A single wrong citation can change the status of an argument. A fake case can do more than embarrass an author. It can mislead editors, courts, reviewers, and readers.

The best known warning remains Mata v. Avianca, where the Southern District of New York sanctioned lawyers after they submitted non-existent cases and fake quotations generated through ChatGPT. The problem did not stop at tool use. The court focused on the failure to verify the cited authorities after the citations came under question. (law.justia.com)

Academic legal writing is not the same as litigation. Still, the lesson travels well.

If AI helps you find cases, summarize doctrine, draft footnotes, translate source material, or shape an argument, you need a record of what the tool did and how you checked it.

Our general guide to ChatGPT disclosure in academic papers covers broad scholarly writing. Legal research needs one more layer: citation and authority verification.

What law reviews now ask authors to disclose

Law review policies now vary. Some ask for a broad disclosure. Others ask only for uses that affected substance, originality, or reliability. Some ask every author to fill out a form whether they used AI or not.

Stanford Law Review states that authors must disclose generative AI use that significantly affected the substance, originality, or reliability of a submission. Stanford also asks authors to complete a form regardless of whether they used generative AI. (stanford-law-review.scholasticahq.com)

The University of Pennsylvania Law Review requires authors, beginning with Volume 175, to disclose any AI use in preparing submissions. Its author page asks for a short description in the cover letter or an addendum that states which tools authors used, how the tools assisted research, and how the tools assisted drafting or revision. It also states that disclosure alone will not count against the submission. (penn-law-review.scholasticahq.com)

NYU Law Review’s 2026 author policy asks authors to complete an AI form or submit a customized disclosure form even when they did not use AI. It defines AI for that policy as generative AI and lists tools such as ChatGPT, Claude, Google Gemini, Midjourney, DALL-E, Sora, and ElevenLabs as examples. The policy also states that AI software may not author or co-author a submission. (nyulawreview.org)

That variety creates a practical problem. You may submit the same article to several journals with different rules.

Do not wait until the upload screen asks a question. Keep one source record during the writing process, then adapt it to each journal. An AI Usage Card gives you that source record.

For journal-level context, see [[AI disclosure](/how-to-disclose-ai-use-for-neurips-icml-and-acl-submissions/) policies by major journals](/ai-disclosure-policies-by-journal/) and AI transparency requirements for journal submissions.

What to record while you work

A useful AI record for legal research should answer six questions.

First, name the tool. Include the version if the interface gives one. If you used a legal AI system, a general chatbot, a citation tool with generative features, or AI inside a word processor, name that system.

Second, state the task. "Used ChatGPT for research" tells an editor almost nothing. "Used Claude to generate possible search terms for state constitutional cases on school funding" gives a clear account.

Third, record the material you entered. Do not paste privileged, confidential, restricted, or personal data into public AI tools. If you used public case text, say so. If you used your own draft, say so. If you used interview excerpts, student records, client material, court filings under seal, or unpublished archive material, stop and check the governing rules before you proceed.

Fourth, state what the tool produced. Did it suggest cases? Summarize doctrine? Draft a paragraph? Rewrite footnotes? Translate a statute? Identify counterarguments?

Fifth, describe your verification. This matters most in law. Did you check every case in Westlaw, Lexis, official court websites, HeinOnline, or another source? Did you confirm quotations against the reporter? Did you review statutory text from the official code? Did you inspect docket entries?

Sixth, say what entered the manuscript. Some AI outputs never make it into a paper. Other outputs shape the structure. Editors need the difference.

You can track this in a simple log, then generate a card when you submit. The template in AI tools for research: a disclosure log template works well for legal projects too.

A simple law review disclosure statement

Use plain language. Do not over-explain. Do not apologize for allowed use.

For a cover letter or addendum, you might write:

I used Claude 3.5 Sonnet in March 2026 to generate search terms for identifying federal appellate cases on qualified immunity and municipal liability. I did not rely on AI-generated citations as authority. I checked all cases, quotations, and pincites in Westlaw before including them in the manuscript. I also used Microsoft Copilot in Word for grammar suggestions during final revision. I accepted or rejected all changes myself.

That statement works because it names the tool, task, stage, and verification step.

If the journal asks for less detail, shorten it:

I used Claude 3.5 Sonnet to generate legal research search terms and Microsoft Copilot for grammar suggestions. I verified all authorities and quotations through standard legal research databases and retained responsibility for the final text.

If the journal asks for a structured form, copy the relevant fields from your AI Usage Card instead of rewriting the disclosure from memory.

For more examples, see [AI Usage Cards examples and templates](/ai-usage-cards-examples/).

When AI helps with citations

Legal scholars should treat AI-generated citations as unverified leads.

A tool may produce a real case with the wrong holding. It may give a correct case with a fake quotation. It may cite a statute section that changed. It may mix jurisdictions. It may invent a reporter citation that looks plausible.

The fix is boring and effective: verify every authority before the manuscript leaves your desk.

A disclosure does not rescue an unchecked citation. It tells the editor what happened. Your verification does the real work.

A strong AI Usage Card for citation work should say something like this:

AI suggested candidate authorities. The author verified existence, jurisdiction, holding, quotation, and pincites against primary or standard legal sources before citing them.

That sentence helps editors see that you did not treat the AI output as law.

When legal ethics rules enter the picture

Some legal scholars also practice law, supervise clinics, or work with client material. For them, disclosure is not only a publication issue.

The American Bar Association released Formal Opinion 512 on July 29, 2024. The ABA described the opinion as its first ethics guidance on lawyers’ use of generative AI tools. The guidance discusses existing duties under the Model Rules, including competence, confidentiality, communication, and reasonable fees. (americanbar.org)

That matters for scholarship when a research project overlaps with representation, clinical work, expert work, or confidential files.

If your article uses client facts, sealed filings, nonpublic discovery, interview material, or clinic documents, do not put that material into a public AI system unless your rules, approvals, and consents allow it. Your AI disclosure should not reveal confidential information either. State the category of use without exposing the protected material.

For example:

I used a locally hosted language model to summarize de-identified interview memos. I did not enter client names, docket numbers, sealed material, or identifying facts into a public AI system. I reviewed all summaries against the original memos.

If you work with human subjects, student records, or sensitive interviews, also read AI disclosure for qualitative research and AI ethics and documentation in academic research.

When student work or law school rules apply

Law school rules can be stricter than journal rules.

Berkeley Law’s AI policy, for example, prohibits AI use for conceptualizing, outlining, drafting, revising, translating, or editing work submitted for credit, and allows AI research on papers only for identifying sources such as cases, statutes, or secondary sources. The same policy says instructors may adopt different written rules and must require disclosure of authorized AI use. (law.berkeley.edu)

A student note can therefore face two rule sets at once: the law school rule and the journal rule.

Follow the stricter rule. If you used AI with instructor permission, keep that permission in writing. Then make the disclosure match both the course policy and the publication policy.

A student note disclosure might say:

Under written instructor permission, I used ChatGPT to identify possible search terms for sources on state consumer privacy statutes. I did not use AI to outline, draft, revise, translate, or edit the note. I verified all sources myself before citation.

Graduate students and doctoral researchers can adapt the guidance in How to disclose AI usage in your thesis when legal research forms part of a dissertation.

Empirical legal studies need a fuller record

Doctrinal work often turns on sources and interpretation. Empirical legal studies add data choices.

If you use AI to classify cases, code judicial opinions, extract variables from filings, summarize interviews, translate responses, generate code, or clean datasets, your disclosure should include method details.

Name the tool. Name the task. Describe the sample. Explain the human review process. State whether AI output became data, analysis, prose, or only a preliminary aid.

For a paper in LaTeX, you can add a short disclosure note like this:

\section*{AI Usage Disclosure}
 
The author used ChatGPT-5 on 2026-04-12 to generate preliminary search terms for identifying federal district court opinions involving algorithmic decision systems. The author did not use AI-generated case citations as authority. All cases, quotations, and pincites were verified in Westlaw before inclusion.
 
The author also used Claude 3.5 Sonnet to draft Python comments for data cleaning scripts. The author reviewed and revised all code before analysis. No confidential, sealed, or personally identifying material was entered into the AI tools.

If you use Overleaf, you can place that section before acknowledgments or in an appendix, depending on the journal format. See How to use AI Usage Cards in Overleaf and the LaTeX tutorial for AI Usage Cards.

Where an AI Usage Card fits in legal scholarship

An AI Usage Card does not replace a journal’s required form. It helps you answer the form without guessing.

For legal scholarship, you can use the card in four places.

You can attach it as a disclosure addendum to a law review submission. You can copy the short text into a cover letter. You can keep the full card in your project archive for editors or co-authors. You can add a shorter version to the acknowledgments or methods section after acceptance.

That last option matters for edited volumes and peer reviewed law journals. Readers may not see your cover letter. A short disclosure in the final manuscript gives them the information they need.

Try this structure:

AI usage: I used [tool] for [task] during [stage]. I verified [legal authorities/data/output] by [method]. I retained responsibility for all analysis, citations, and final text.

If you used no AI, some journals may still ask you to say that. In that case:

AI usage: I did not use generative AI tools to research, draft, revise, translate, or edit this manuscript.

Generate the record before submission day

Submission day invites bad disclosure. You are tired. The article has too many footnotes. Scholastica asks for a statement, and you try to remember what happened six months ago.

Do not rely on memory.

Generate an AI Usage Card at ai-cards.org when you start using AI in a legal research project. Update it before submission. Then copy the short disclosure into the cover letter, attach the card as an addendum if the journal allows it, or add the text to your manuscript.

Legal scholarship runs on trust in sources. Your AI disclosure should help readers see how you earned that trust.