Same School, Different AI Rules: When Programs Contradict Each Other

Same School, Different AI Rules: Program-Level Policy Contradictions That Could Cost You Admission

Here's a scenario that trips up applicants every cycle: you look up whether University X allows AI in applications. You find a permissive policy. You use AI for brainstorming. Then your application gets flagged because the specific program you applied to has a completely different, much stricter policy that you never saw.

This isn't hypothetical. At least five major universities maintain AI policies that directly contradict each other across programs. The institution says one thing. The law school says another. The business school says something else entirely. And you, the applicant, are expected to find the right one.

Here are the most dramatic contradictions we've found, verified against official policy documents and program pages.

University of Michigan: AI Required for One Essay, Restricted for Others

Undergraduate admissions: L3/D0/E0 (assistance permitted, authorship reserved)

Michigan's undergraduate admissions page now publishes its own guidance, read on September 1, 2026:

"AI tools can be helpful when used thoughtfully, such as for brainstorming, outlining, proofreading, or organizing information."

"your application should reflect your own ideas, experiences, and authentic voice. Do not copy and paste AI-generated text or submit work that does not represent you."

That is a rule about assistance, not a ban: brainstorming, outlining and proofreading are named as acceptable, and pasted AI text is not. No disclosure requirement, and no published detection method. (An earlier version of this page said Michigan simply pointed at the Common App fraud policy; that is no longer the published position, and we corrected it on September 1, 2026.)

But drill into the other units and the picture changes completely.

Rackham Graduate School: L3/D3/E0

The Rackham Graduate School permits limited AI use but demands formal attestation (read September 1, 2026):

"the Rackham Graduate School application will require your attestation that you are the sole author of all written passages in your application"

"GenAI tools, like Grammarly, may be used to review grammar and spelling of your application essays"

You can use Grammarly for grammar and spelling. You may not use GenAI "to outline, substantively draft, or write the content of your application essays", and admission "may be revoked if any part of the application contains falsifications, misrepresentations, or omissions". Rackham publishes no detection or screening method, so its enforcement code is E0. It is part of a broader pattern we cover in the Grammarly question no college will answer.

Law School: L4/D0/E0, with one L1 exception (The Paradox)

Michigan Law takes the most unusual approach of any school in our database. Its 2026-27 application page reads (checked September 1, 2026):

"Please note that for all written submissions, we expect that the work is the applicant's own, meaning that the ideas and expressions originated with the applicant, and that the applicant wrote all drafts and the final product. In general, applicants ought not use generative artificial intelligence tools as part of their drafting process. (There is one exception to this rule, however, discussed below under 'Supplemental Essays.')"

That exception is optional supplemental essay 10:

"TO BE ANSWERED USING GENERATIVE AI: Discuss how you have used AI to solve a real-world problem. How was AI helpful, and how was it lacking? Going forward, how will you approach the use of such tools?"

Read that again. Michigan Law requires generative AI for one specific optional essay while telling applicants they ought not use it anywhere in their drafting. The page asks you to be comfortable saying "yes" if asked whether the essay is your own work; the certification you actually sign sits inside the LSAC application, and we have not read its 2026-27 text. You need to track, essay by essay, which ones permit AI and which ones forbid it.

The practical problem: A student applying to Michigan's undergraduate program, Rackham, Ross and the Law School simultaneously faces four different AI frameworks at the same institution. Get confused about which policy applies to which essay, and you've either violated a prohibition or failed to use a required tool.

View Michigan's full policy breakdown

UW-Madison: "We Won't Disqualify You" vs "We'll Rescind Your Admission"

Institution Level: L1/D0/E0 (explicitly permissive)

UW-Madison's undergraduate admissions page contains one of the most applicant-friendly AI statements in higher education:

"We will not disqualify an applicant found to have used or suspected of using AI in their admissions essays, nor are we running essays through any system to detect if AI was used."

They also add:

"However, we strongly discourage students from simply feeding AI a prompt for their essay, as it denies you the opportunity to help us better understand why you would be a good fit for UW-Madison."

This is about as permissive as it gets. No disqualification. No detection systems. No disclosure requirement. Just a gentle discouragement.

La Follette School of Public Affairs: L4/D3/E3 (Polar Opposite)

Now read what the La Follette School says to its graduate applicants:

"We require you to certify that the essays you submit are entirely your own original work and have not been created using Artificial Intelligence (AI). Please be aware that if it is discovered that the submitted essays do not meet this requirement, we retain the right to revoke your admission."

That's a total prohibition (L4), formal certification requirement (D3), and active enforcement with admission rescission (E3). It is the maximum possible restriction on every dimension.

The whiplash: UW-Madison's undergraduate admissions explicitly says they won't even look for AI. The La Follette School at the same university says they'll revoke your admission if they find it. The institutional policy and the program policy aren't just different. They're diametrically opposed.

View UW-Madison's full policy breakdown

Columbia University: Institutional Silence, Program-Level Severity

Institution Level: L0/D0/E0 (no policy found)

Columbia University has no university-wide AI policy for admissions. Their institutional page effectively says:

"No official AI policy."

But three of Columbia's programs have taken aggressive independent positions.

School of the Arts (Graduate): L4/D3/E3

"Forbids the use of any generative artificial intelligence platforms in the creation of application materials."

Total prohibition with attestation and active enforcement. If you're applying to Columbia's MFA programs, AI is completely off limits.

Graduate School of Arts and Sciences: L4/D0/E2

"Artificial intelligence programs may not be used to produce any application materials or responses."

Another total prohibition, this time with screening-based enforcement but no formal disclosure requirement. The school assumes you'll comply and checks rather than making you attest.

Business School (MBA): L2/D0/E0, reported

Columbia Business School's own MBA application-requirements page returned an error to all three retrieval routes we tried on September 1, 2026, so we have no first-party text. A third-party admissions guide (page updated July 2026) reports the rule as permitting generative AI "for idea generation and/or to edit a candidate's work" while treating a complete AI-generated response as an Honor Code violation. That is reported, not published — read the wording inside the application before relying on it. An earlier version of this section quoted an "exclusively your own" sentence as Columbia Business School's and called its enforcement active; neither was sourced, and we removed both on September 1, 2026.

The gap: An applicant to Columbia's undergraduate program finds zero guidance. An applicant to the School of the Arts at the same university finds one of the strictest policies in the country. If you're applying to multiple Columbia programs, you need to check each one individually.

View Columbia's full policy breakdown

University of Pennsylvania: Institution Allows, Wharton and Law Forbid

Institution Level: L1/D1/E1 (limited AI permitted)

Penn's institutional AI guidance states:

"If it is unacceptable to have another person substantially complete a task, it is also unacceptable to have AI complete the task."

And:

"Be transparent about the use of AI. Disclose when a work product was created wholly or partially using an AI tool."

This allows brainstorming, idea generation, and limited assistance. It asks for disclosure. It's a reasonable middle-ground framework.

Wharton School (MBA): L0/D0/E0

Wharton publishes no AI text at all. We read the MBA application guide, its essays section and the admissions FAQ on September 1, 2026 and found no AI, authorship, plagiarism or detection rule on any of them.

A sentence saying Wharton "may use its own proprietary and/or licensed AI solutions in order to identify AI-authored elements of applications" circulates in third-party guides (reported, May 2026). It is not on any Wharton page, and "may use" is not the same as detecting AI — treat it as reported and confirm inside the application. An earlier version of this section described the sentence as a fact and called Wharton a total prohibition with active screening; we corrected that on September 1, 2026.

Penn Law School: L4/D3/E2

Penn Law goes even further:

"Work must be sole and original."

Total prohibition with a formal attestation requirement. You must certify that your work is entirely your own, and the school actively screens submissions.

The contradiction: Penn's institutional policy explicitly says "be transparent about the use of AI," implying it's acceptable if disclosed. Penn Law says your work must be sole and original, and Wharton publishes no admissions AI rule at all. A dual-degree applicant to Penn's main programs and Wharton would face two contradictory frameworks from the same university.

View Penn's full policy breakdown

Stanford University: Institutional Ambiguity vs GSB Absolutism

Institution Level: L0/D1/E1 (no specific policy)

Stanford's institutional position is minimal. Their guidance emphasizes authenticity:

"Essays should allow the applicant's genuine voice to come through."

No specific mention of AI prohibition. No disclosure framework. Just an emphasis on voice and authenticity. Stanford's AI Advisory Committee has focused on how the university itself uses AI, not on restricting applicants.

Graduate School of Business: L4/D0/E3

Stanford GSB takes one of the hardest lines of any business school:

"It is improper to have another person or tool write your essays. Such behavior will result in denial."

Not rescission. Not review. Denial. The GSB treats AI essay assistance as grounds for automatic rejection, with the strongest possible enforcement. No disclosure process because there's nothing to disclose: the answer is simply no.

The disconnect: A student applying to Stanford's graduate engineering program finds no AI guidance whatsoever. The same student adding a GSB application faces one of the most punitive policies in higher education. At Stanford, the question isn't "does the university allow AI?" It's "which program are you applying to?"

View Stanford's full policy breakdown

Why This Matters More Than You Think

These contradictions aren't academic curiosities. They have real consequences for applicants:

1. Dual-degree applicants are navigating a minefield. If you're applying to Penn's general graduate programs and Wharton simultaneously, you need two different AI strategies for the same university.

2. "Does University X allow AI?" is the wrong question. The right question is always "does this specific program at University X allow AI?" The institution-level policy may have zero bearing on your program.

3. Program-specific policies override institutional policies. In every case we've documented, the program-level policy is stricter than the institutional policy. Never assume the institutional policy is the ceiling. It's usually the floor. The schools with the most detailed AI admissions policies tend to be the same ones with the deepest program-level fragmentation -- the more a school writes, the more its individual schools tend to deviate.

4. Silence isn't permission. When Columbia's institution has no policy but Arts has an L4 prohibition, the absence of a university-level policy doesn't mean Arts applicants can use AI. It means the university hasn't caught up to what its programs are already doing.

How to Navigate This

1. Always check the specific program's admissions page. Don't stop at the university-level policy. Look for the admissions requirements for your exact program, department, or school.

2. Check GradPilot's AI policy database. We track both institution-level and program-specific policies, so you can see exactly where contradictions exist.

3. Read our methodology page to understand how program-specific overrides work in our L/D/E rating system.

4. When program and institution policies conflict, follow the stricter one. You'll never get in trouble for being too cautious.

5. If you're applying to multiple programs at the same university, treat each application as having its own AI rules. Because it does.

6. Contact the admissions office directly if you're unsure. A quick email asking "does your program have specific AI guidelines for application essays?" can save you enormous stress.

The age of "one school, one policy" is over. Programs are setting their own rules, often without coordinating with the institution or each other. The applicants who succeed will be the ones who check every program individually.

Related Reading

• Canadian Application Feedback Rules: Where Help Stops
• Georgetown vs Caltech: Two Models for AI in Admissions
• Do Top 10 Colleges Check for AI? Official Policies
• Do Colleges Use AI Detectors? The Truth About Turnitin
• Colleges Ban Student AI but Use AI to Read Essays
• AI Reads Your College Essays: Virginia Tech, UCLA, Penn State