Can You Use AI on a Law School Essay? It Depends on the Verb

Can You Use AI on a Law School Essay? It Depends on the Verb

Two applicants run the same finished draft through the same tool on the same afternoon. One is inside their school's published rule and one is outside it. The difference is not that one school is stricter. It is that the two schools wrote sentences about different things: one prohibited assistance in completing the application, the other prohibited drafting.

That is why there is no market-wide answer to "can I use AI on my law school personal statement." We read sixteen schools' own pages on July 29, 2026 and sorted them by the object each sentence regulates rather than by how strict it sounds:

• The application itself. Any tool touching any part of it is inside the ban. UCLA, Columbia, Buffalo.
• The drafting. Where the words originated — usually with an explicit carve-out for proofreading or feedback. Michigan, Duke, Baylor, Georgetown, UVA.
• Editing, affirmatively permitted with conditions. Enumerated allowances, in one case with required disclosure. Richmond, USC Gould, ASU.
• Nothing published. No first-year J.D. essay AI rule on the school's own requirements pages. Harvard, Yale, Berkeley, Texas, Temple.

And for that fourth group, the sentence that has to travel with it: absence is not permission.

Why there is no single rule

Law applicants keep hunting for the platform-level answer because every other part of the process has one. LSAC carries your transcripts, your letters and your report — but not an essay policy. The LSAC Candidate Agreement for 2026-27 bars giving, receiving or requesting assistance during the LSAT and LSAT Argumentative Writing. It is a test document, scoped to the test, and nothing in it establishes what you may do with an essay you write for a school months later.

LSAC did publish a view, once, on its own blog in 2023: "In short, a complete ban of ChatGPT would be challenging to enforce and justify." The same post described testing AI-detection products and being "scandalized when one tool repeatedly and erroneously identified my own work as likely written by ChatGPT," concluding that "these detection tools did not appear to meet that high standard." A blog post is not a policy, and LSAC has not made it one.

Compare the medical side, where the application platforms are the policy sponsors: AMCAS, CASPA and TMDSAS each publish their own AI rule, and every school on the platform inherits it. In law, the platform carries the file and the school writes the rule — so the rule surface is every school drafting its own sentence on its own timeline.

The four things a law school AI policy can regulate

Class A: the object is completing the application

The broadest class. The verb is "assist in the completion," and it does not care whether the tool produces text. UCLA School of Law states it in one line. Columbia Law's most recent published preview application — the Fall 2026 one, since the Fall 2027 preview was not yet published on July 29, 2026 — has applicants certify that no AI tool or service "has written or modified any component of the application." University at Buffalo School of Law is the only school in our sample that names editing inside the prohibition itself.

This is the one class where an outside review tool is unambiguously excluded: on its face, a review is assistance in completing the application.

Class B: the object is drafting

Here the regulated thing is where the words came from, and several of these schools tell you in the same paragraph what they did not ban. Georgetown Law draws it most cleanly: "While you can (and should) receive and incorporate feedback on drafts, the only person who may be engaged in the actual writing is you." Duke Law bars AI from all drafts and the final product, then adds that this "does not preclude asking family members, friends, pre-law advisors, and others for proofreading assistance or general feedback." Baylor Law uses the narrowest verb of all — write — in a sentence that begins by permitting a pre-law advisor to review the statement. UVA Law asks for a statement written "without the help of artificial intelligence tools," which reaches help, not just authorship.

This is the class most often mistaken for a blanket ban. Whether a review of your finished text sits inside it depends entirely on the carve-out wording, and the carve-outs are not the same sentence.

Class C: editing permitted, with conditions

Three schools affirmatively permit some AI use and enumerate it. Richmond Law publishes a permitted list and a not-permitted list, and requires disclosure. USC Gould permits tools "for editing or guidance" while stating that "the final content should authentically reflect you," with no disclosure requirement attached. ASU's Sandra Day O'Connor College of Law announced in July 2023 — still live today — that applicants may use generative AI "in the preparation of their application" and certify that what they submit is accurate.

Class D: nothing published

Five schools in our sample publish no first-year J.D. essay AI rule on their own requirements pages. Four of them are among the best-known law schools in the country. What that means, and what it does not, gets its own section below.

The sixteen sentences, sorted by what they govern

Every row was fetched and read on July 29, 2026. Quotes are the schools' own words; the class is ours.

Sixteen schools is a sample, not a census, and no row here should be read as a claim about law schools generally. If your school is not listed, look it up: we keep a school-by-school directory of published admissions AI policies at the law school AI policy database — that page is the lookup surface, this post is how to read what you find on it.

One disclosure about that directory, because it is our own evidence for the point this post is making: it is being corrected in the same release as this post. Some law cards had been built from institution-level or LL.M.-scoped policies, which is how a couple of schools that publish nothing for J.D. essays ended up displayed as though they had bans. The corrected rows carry the same July 29, 2026 check dates as the table above.

The school that bans AI in drafting and requires it in one essay

Michigan Law is the sharpest case in the sample. Its J.D. application page sets the drafting rule, then names the carve-out in the next breath: "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." Its own summary test: "At the end of the day, you should be comfortable saying 'yes' if asked the question, 'Is this essay your own work?'"

Then the exception. Among Michigan's optional supplemental prompts is one where generative AI is not merely allowed but required: "Essay Ten – TO BE ANSWERED USING GENERATIVE AI: How much do you use generative AI tools such as ChatGPT right now? What's your prediction for how much you will use them by the time you graduate from law school? Why?" The drafting rule holds for every other prompt; answer this one and you must use AI to do it.

A caveat on the certification: Michigan's most recent published annotated application is dated June 2025, for the 2025-26 season. In it the applicant affirms that nobody assisted beyond basic proofreading or a general critique, that generative AI was not used in drafting except for that one optional prompt, and that false statements may result in revocation of an offer of admission, expulsion, or the rescinding of a degree. The 2027 certification is not public yet, so read the one inside your own application.

The school that permits editing and requires you to say so

Richmond Law, on its Fall 2027 J.D. application checklist, is the counterexample to every headline that says law schools ban AI. It publishes two lists.

Not permitted: "Using generative AI to draft your personal statement, addenda, short answers, or other written components of the application in lieu of your own original work," and "Using AI to substantially re-write your work."

Permitted: using AI tools to develop and select a theme for the personal statement; "Using AI tools to check spelling, grammar, and word choice, including in your personal statement or other narrative submissions and short answer questions;" and "Editing your original writing with AI tools to assist with length or clarity, while preserving your ideas, structure, and substantial content."

Two details are easy to miss. First, the page states that an applicant who uses AI tools must disclose "the tools used and the way you used them" — a requirement, and no other school in our sample publishes one. Second, Richmond's own definition of "AI tools" is broad: natural language processing algorithms, machine learning algorithms, large language models and generative AI, with ordinary grammar-checking services named alongside chat assistants. Under that definition, a spellchecker is an AI tool you would be disclosing.

Absence is not permission

Four of the most-searched law schools in the country — Harvard, Yale, Berkeley and Texas — published no first-year J.D. essay AI rule that we could find on their own requirements pages on July 29, 2026. Temple published a permission for human readers and said nothing about AI at all. If you came here to learn whether Harvard Law bans AI on the personal statement, that is the honest answer, and we are not going to convert it into a yes.

A school's public requirements page is not the whole rule surface. The certification you sign inside the application, and any instruction inside the portal, can say something the public page does not, and the certification is usually where the wording is tightest. Michigan is the worked example: its public page and its certification say compatible things, and the certification is the stricter document. Silence on a public page is evidence about the page, not about the school.

So the only instruction here is procedural: read the certification text inside your own application before you submit, and treat that as the governing sentence. We will not tell you what it means for your situation, and you should be wary of anyone who does — including pages that say a school with no published policy is fine, or that some level of use is the safe one. Neither claim has a source. It is the same pattern we found across undergraduate admissions, where most colleges publish no AI policy at all, and the same reason the attestation you sign is worth reading slowly.

Getting feedback without breaking the rule

Several schools that restrict AI affirmatively permit a human reader, in writing. Georgetown says you "can (and should) receive and incorporate feedback on drafts." Michigan invites applicants to ask "pre-law advisors, mentors, friends, or others for basic proofreading assistance and general feedback and critiques." Duke's rule "does not preclude asking family members, friends, pre-law advisors, and others for proofreading assistance or general feedback." Baylor allows that "others (e.g., your pre-law advisor) may review it and offer guidance." Temple welcomes you "to ask family, friends, or advisors for feedback." Five schools, putting a reader inside the rule on purpose.

Georgetown also publishes the only usable self-test we found, and it is about disclosure comfort rather than tools: reflect "on how you would feel being completely open and honest with a member of the Admissions Committee about your process. If you would omit or downplay how you used an AI tool, or how much wordsmithing a trusted friend did in your draft, you should adjust your plan." That is Georgetown's test, not ours, and worth taking literally rather than extending.

But "ask your pre-law advisor" is not available to everyone. LSAC's 2025 Prelaw Questionnaire report found that 20% of test takers had no one to rely on for application advice, with the share differing by race, first-generation status and Pell status. Its 2024-25 Applicant Profile found that more than half of respondents spent over 20 hours on the personal statement alone, against a median of 46 hours across six core components.

That gap is what our law school personal statement review is for. It reads writing you already wrote: it does not draft, does not rewrite, does not hand back replacement prose you could paste in, and part of what it returns is a flag on passages that read as AI-sounding, so you can hear what a reader would hear. A Full Review is $5 and typically comes back in about two to three minutes; two free Quick Reviews a day come with the account. Mechanics are covered separately in how a law school personal statement review works.

There are also schools where you should not use it, and we would rather say so here than have you work it out from a certification page. Where the published sentence reaches assistance in completing the application — UCLA, Columbia and Buffalo in this sample — an outside review sits inside the rule. UVA asks for a statement written without the help of AI tools. Michigan carves out proofreading and critique, but its current-cycle certification is not public, so we treat it the same way. Do not use this, or anything like it, on those schools' essays.

Wherever you are applying: review that school's current authorship, AI and assistance rules before you use any review; do not use one where AI review or outside assistance is prohibited; and submit only your own writing.

What to do before you submit

1. Find your school's own sentence. Not a summary, not a roundup — the sentence, on the school's admissions page. The directory linked above is the fastest way in; the school's page is the source.
2. Name the object it regulates. Completing the application, drafting, editing-with-conditions, or nothing published. That determines whether the thing you want to do is inside or outside the rule. Sorting schools by how strict they sound will not.
3. Read the certification text inside your own application. It is the sentence you are actually signing, it can be tighter than the public page, and at some schools it is the only place the rule appears in full.
4. Re-check late in the cycle. Two of the sixteen above — Richmond and USC Gould — now publish something materially different from what our own dataset recorded in February, and both moved toward permission, which is not the direction the internet assumes. A rule you read in September may not be the rule you certify to in January.

The short version

There is no clean answer to "can I use AI on a law school personal statement" because the question contains sixteen different objects and no common sponsor. LSAC governs the test; the school governs the essay; the certification governs you. Find the sentence, name the noun it regulates, read what you are signing, and check again before you submit. If you want to know what a reader hears in your own draft, the six law rubrics are at /rubrics/law, the law statement review is there when your school's rule leaves room for it, and the lookup lives in the law school AI policy database. Nothing here is legal advice, and nothing here predicts what any school will do — only what each one wrote down, and when we checked.