What Law School AI Rules Prohibit Versus Permit, Rule by Rule

What law school ai rules prohibit versus permit, in one line: all eight prohibitions in our verified spine reach AI writing the essay, only Columbia reaches editing, only Michigan names permitted tools by brand, and no school addresses brainstorming at all. Verified July 2026.

What do law school AI rules actually prohibit?

Every prohibition in our verified spine reaches AI writing the essay, and almost nothing beyond that is common ground. The differences are where the practical answers live, and eight of the 9 schools with rules prohibit AI writing: Columbia, Michigan, Georgetown, Duke, UVA, George Mason, UCLA and Brooklyn Law, each on its own pages, retrieved July 31. Verified July 2026.

QuestionAnswer across the nineDetailAI writing the essayProhibited at eight of nineUSC permits AI for editing or guidance insteadAI editing or modifyingExpressly prohibited at Columbia, expressly permitted at USCThe other seven are silent on editingGrammar and spell checkExpressly permitted at Michigan onlyMichigan names Grammarly and spellcheckBrainstorming with AINo school addresses itGeorgetown expressly declines to draw category linesHuman proofreadingExpressly permitted at fourGeorge Mason, Michigan, Duke and GeorgetownA stated consequenceColumbia and Brooklyn LawRescission or revocation, and dismissal if enrolledAn AI attestationColumbia and MichiganBoth require an affirmative certificationAI requiredMichigan, for one optional promptEssay Ten is labelled to be answered using generative AI

Eight of the 9 schools with rules prohibit AI writing: Columbia, Michigan, Georgetown, Duke, UVA, George Mason, UCLA and Brooklyn Law, each on its own pages, retrieved July 31, 2026. Verified July 2026.

Two of the eight reach further than the personal statement. Columbia's covers any component of the application and UCLA's covers completion of the application, both from their own requirements documents (https://www.law.columbia.edu/sites/default/files/2025-11/jd-preview-application-fall-2026.pdf and https://law.ucla.edu/admissions/jd-admissions/first-year-jd-applicants), retrieved July 31, 2026. Verified July 2026.

Duke's reaches all writing samples and all drafts rather than the final version alone, verbatim: “This means that the ideas and expressions originated with you, and you wrote all drafts and the final product without the assistance of generative artificial intelligence.â€? (https://law.duke.edu/apply/degreeprograms/jd), retrieved July 31, 2026. Verified July 2026.

Which rule covers editing as well as writing?

Columbia's is the only rule in our verified spine that expressly reaches AI modifying an essay rather than writing it, and USC's is the only one that expressly permits editing. The seven schools between those two poles are silent. Verified July 2026.

Columbia's certification uses the words written or modified, verbatim: “I further certify that the application I have provided is exclusively my own work and that no agent, third-party vendor, educational consultant, or artificial intelligence tool or service has written or modified any component of the application.â€? (https://www.law.columbia.edu/sites/default/files/2025-11/jd-preview-application-fall-2026.pdf), retrieved July 31, 2026. Verified July 2026.

USC runs the opposite way, verbatim: “While you may use tools (including artificial intelligence) for editing or guidance, the final content should authentically reflect you.â€? (https://gould.usc.edu/academics/degrees/jd/application/), retrieved July 31, 2026. Verified July 2026.

The seven schools between those poles say nothing about editing, which means a grammar pass sits in undefined territory at most of your list. Verified July 2026.

What do law school AI rules permit?

Two things are expressly permitted anywhere in our spine: human proofreading at four schools, and electronic grammar tools at Michigan alone. Everything else is unaddressed, and georgetown permits feedback on drafts expressly, and George Mason and Duke both permit human proofreading, each on its own pages, retrieved July 31, 2026. Verified July 2026.

Michigan names tools, verbatim: “Applicants are nonetheless welcome to use electronic tools (like Grammarly or spellcheck), or may, similarly, ask pre-law advisors, mentors, friends, or others for basic proofreading assistance and general feedback and critiques.â€? (https://michigan.law.umich.edu/admissions/apply-jd-program), retrieved July 31, 2026. Verified July 2026.

Georgetown permits feedback on drafts expressly, and George Mason and Duke both permit human proofreading, each on its own pages, retrieved July 31, 2026. Verified July 2026.

Note what those permissions are about. Three of the four name people rather than tools, so they do not extend to software however similar the task. Verified July 2026.

Does any rule address brainstorming or outlining?

No school in our verified spine addresses brainstorming, outlining or research assistance as distinct from writing. That silence is the largest single gap in this area, and not one of the 9 rules names brainstorming, outlining or idea generation, retrieved July 31, 2026. Verified July 2026.

Not one of the 9 rules names brainstorming, outlining or idea generation, retrieved July 31, 2026. Verified July 2026.

Georgetown is the only school to acknowledge the gap, saying it cannot provide specific guidance on every scenario and offering a disclosure test instead (https://www.law.georgetown.edu/admissions-aid/jd-admissions/first-year-applicants/frequently-asked-questions/), retrieved July 31, 2026. Verified July 2026.

Michigan's wording is the only one that arguably leaves pre-drafting outside the rule, because it is scoped to the drafting process by its own words (https://michigan.law.umich.edu/admissions/apply-jd-program), retrieved July 31, 2026. That is our reading rather than Michigan's statement. Verified July 2026.

Which rules state a consequence?

Two of the nine state a consequence. Seven state none, which leaves applicants following a rule without knowing what breaking it costs, and columbia's, verbatim: “If dishonesty in this regard is discovered, it is grounds for immediate rescission of an admission offer and, if already enrolled, disciplinary proceedings.â€?, retrieved July 31, 2026. Verified July 2026.

Columbia's, verbatim: “If dishonesty in this regard is discovered, it is grounds for immediate rescission of an admission offer and, if already enrolled, disciplinary proceedings.â€? (https://www.law.columbia.edu/sites/default/files/2025-11/jd-preview-application-fall-2026.pdf), retrieved July 31, 2026. Verified July 2026.

Brooklyn Law's, verbatim: “Brooklyn Law School reserves the right to revoke admission or dismiss a matriculated student if any information provided, including in the application, essays, correspondence, or other documents, is found to be generated with assistance from automated intelligence tools.â€? (https://www.brooklaw.edu/media/ifcp4vl1/2026-brooklyn-law-school-application-instructions.pdf), retrieved July 31, 2026. Verified July 2026.

The other seven attach nothing to the AI rule, though most carry a general false-information clause elsewhere that an office could invoke, retrieved July 31, 2026. Verified July 2026.

Which rules require an attestation?

Columbia and Michigan require an affirmative certification about AI. That converts the rule from an instruction into something you have signed, which changes its weight entirely, and michigan's carries its own exception, verbatim: “I certify that the thoughts and words contained in my personal statement and essays are my own. Verified July 2026.

Columbia's, verbatim: “I further certify that the application I have provided is exclusively my own work and that no agent, third-party vendor, educational consultant, or artificial intelligence tool or service has written or modified any component of the application.â€? (https://www.law.columbia.edu/sites/default/files/2025-11/jd-preview-application-fall-2026.pdf), retrieved July 31, 2026. Verified July 2026.

Michigan's carries its own exception, verbatim: “I certify that the thoughts and words contained in my personal statement and essays are my own. Likewise, I have not used generative artificial intelligence tools as part of my drafting process, unless I answered optional essay prompt 10.â€? (https://michigan.law.umich.edu/admissions/apply-jd-program), retrieved July 31, 2026. Verified July 2026.

The other seven publish no AI attestation, so their rules operate as instructions. Breaching an instruction is a policy matter; breaching a certification is a candour matter that reaches bar admission. Verified July 2026.

How strong is the language, school by school?

The nine rules use four different registers, from strictly prohibited to should not to a flat permission. Reading the modal verb tells you how the school expects to enforce it, and columbia says strictly prohibited and George Mason says prohibited, both unqualified, from their own pages, retrieved July 31, 2026. Verified July 2026.

Columbia says strictly prohibited and George Mason says prohibited, both unqualified, from their own pages (https://www.law.columbia.edu/sites/default/files/2025-11/jd-preview-application-fall-2026.pdf and https://www.law.gmu.edu/admissions/jd/how_apply_jd/), retrieved July 31, 2026. Verified July 2026.

UVA uses should rather than must, verbatim: “The statement should be written in your own voice without the help of artificial intelligence tools and should not exceed two pages double-spaced in 12 pt font.â€? (https://www.law.virginia.edu/admissions/admissions/jd-application-information), retrieved July 31, 2026, and Michigan uses ought not with an in general qualifier. Verified July 2026.

USC uses may, in a permission, verbatim: “While you may use tools (including artificial intelligence) for editing or guidance, the final content should authentically reflect you.â€? (https://gould.usc.edu/academics/degrees/jd/application/), retrieved July 31, 2026. That is the full range published anywhere in our spine. Verified July 2026.

Does LSAC ban AI, and how is that different?

LSAC does publish an AI prohibition, and it governs LSAT Argumentative Writing rather than the personal statement. Confusing the two is the commonest error in this area, and the consequences differ sharply, and you may not use any form of generative AI program to assist you while taking LSAT Argumentative Writing.�, retrieved July 31, 2026. Verified July 2026.

LSAC's answer, verbatim: “No. You may not use any form of generative AI program to assist you while taking LSAT Argumentative Writing.â€? (https://www.lsac.org/lsat/frequently-asked-questions-about-lsat/frequently-asked-questions-about-lsat-argumentative), retrieved July 31, 2026. Verified July 2026.

LSAC extends that to grammar tools, verbatim: “Before launching LSAT Argumentative Writing, be sure to shut down all non-essential applications and disable any grammar-checking extensions, including Grammarly, Microsoft Editor, etc.â€? (https://www.lsac.org/lsat/frequently-asked-questions-about-lsat/frequently-asked-questions-about-lsat-argumentative), retrieved July 31, 2026. Verified July 2026.

LSAC states a consequence where most schools do not, verbatim: “Use of these types of program will result in the cancellation of your LSAT Argumentative Writing and may also result in a test security investigation and/or a referral to LSAC's Subcommittee on Misconduct and Irregularities in the Admission Process.â€? (https://www.lsac.org/lsat/frequently-asked-questions-about-lsat/frequently-asked-questions-about-lsat-argumentative), retrieved July 31, 2026. Verified July 2026.

One carve-out is published: the spellcheck built into the LawHub interface is allowed while external AI-assistive extensions are not (https://www.lsac.org/lsat/frequently-asked-questions-about-lsat/frequently-asked-questions-about-lsat-argumentative), retrieved July 31, 2026. LSAC's own JD application requirements page carries no AI rule at all. Verified July 2026.

What should you do when your schools disagree?

Work to the strictest rule on your own list, because you submit one personal statement to all of them through LSAC. That is the only coherent answer and we will not hedge it, and an essay edited with AI is compliant at USC and non-compliant at Columbia, and you are sending the same document to both. Verified July 2026.

If Columbia is on your list, its rule governs your essay everywhere, because its certification reaches any component written or modified by AI (https://www.law.columbia.edu/sites/default/files/2025-11/jd-preview-application-fall-2026.pdf), retrieved July 31, 2026. Verified July 2026.

USC's permission does not travel. An essay edited with AI is compliant at USC and non-compliant at Columbia, and you are sending the same document to both. Verified July 2026.

The practical consequence is that the most permissive school on your list is irrelevant to your drafting decisions. Only the strictest one matters. Verified July 2026.

Which schools does this cluster cover, and which does it not?

The sweep behind this page covered 59 law schools in our verified spine, of which 9 publish an AI rule and 50 were checked and publish none. There are roughly 200 ABA-accredited law schools, and every school was checked on its own admissions pages, application requirements, admissions FAQ and school-hosted application PDFs on July 31, 2026. Verified July 2026.

Every school was checked on its own admissions pages, application requirements, admissions FAQ and school-hosted application PDFs on July 31, 2026. No aggregator, prep site or ABA portal was used. Verified July 2026.

Absence from the rule list means checked and none published rather than unregulated. Every school still requires a general certification that the application is accurate, and a school without an AI clause can act on a false certification. Verified July 2026.

Nine pages we could not read are recorded rather than glossed, including Fordham's application requirements page and FAQ behind an SSO gateway and Wake Forest's certification panel, which renders only in JavaScript. Those schools are none-found rather than cleared. Verified July 2026.

Where every figure on this page came from

Every figure on this the scope of law school AI rules page carries a Verified July 2026 stamp because a dated stamp is the only honest way to publish a number that moves. Each source below was retrieved in July 2026 and is linked in full rather than named vaguely, so nothing here is modelled, averaged or inferred.

What it supportsSourceColumbia AI rule, application form/PDFhttps://www.law.columbia.edu/sites/default/files/2025-11/jd-preview-application-fall-2026.pdfMichigan AI rule, requirements page and application PDFhttps://michigan.law.umich.edu/admissions/apply-jd-programGeorgetown AI rule, FAQhttps://www.law.georgetown.edu/admissions-aid/jd-admissions/first-year-applicants/frequently-asked-questions/Duke AI rule, requirements pagehttps://law.duke.edu/apply/degreeprograms/jdUVA AI rule, requirements pagehttps://www.law.virginia.edu/admissions/admissions/jd-application-informationGeorge Mason AI rule, requirements pagehttps://www.law.gmu.edu/admissions/jd/how_apply_jd/UCLA AI rule, requirements pagehttps://law.ucla.edu/admissions/jd-admissions/first-year-jd-applicantsBrooklyn Law AI rule, application form/PDFhttps://www.brooklaw.edu/media/ifcp4vl1/2026-brooklyn-law-school-application-instructions.pdfUSC AI rule, requirements pagehttps://gould.usc.edu/academics/degrees/jd/application/LSAC on AI in LSAT Argumentative Writinghttps://www.lsac.org/lsat/frequently-asked-questions-about-lsat/frequently-asked-questions-about-lsat-argumentative

FAQ

What do law school AI rules prohibit?

All eight prohibitions in our spine reach AI writing the essay. Columbia's and UCLA's reach the whole application, and Duke's reaches all drafts rather than the final version alone. Verified July 2026.

What do law school AI rules permit?

Human proofreading at four schools, and electronic grammar tools at Michigan alone, which names Grammarly and spellcheck. USC permits AI for editing or guidance. Verified July 2026.

Does any rule cover AI editing?

Columbia's does, expressly, through a certification covering work written or modified by AI. USC expressly permits editing.

The other seven schools are silent. Verified July 2026.

Do the rules address brainstorming?

No school in our verified spine addresses brainstorming or outlining. Georgetown acknowledges the gap and offers a disclosure test rather than a category list. Verified July 2026.

Which rule should I follow if my schools disagree?

The strictest one on your list, because you submit one personal statement to all of them. USC's permission does not travel to Columbia. Verified July 2026.

What to do next

Nine rules that agree on one thing and diverge on everything else. The strictest one on your own list is the only one that changes what you should do. The Lovare Diagnostic takes about fifteen minutes and returns your own numbers against published school data rather than a single probability, and it is free.

Written by Ali, Georgetown Law, founder of Lovare Institut.

Book a call