NESCAC's 12 Liberal Arts Colleges Have 1 AI Policy

NESCAC's 12 Liberal Arts Colleges Have 1 AI Policy Between Them

A cohort of small, writing-obsessed New England liberal arts colleges — schools whose entire pedagogical identity is built around the seminar table, the handwritten essay comment, and the honor code — has produced almost nothing in writing about AI in admissions.

We pulled the New England Small College Athletic Conference (NESCAC) and its closest peer, UMass Amherst, out of our 174-school AI policy dataset. The result: 12 schools, 10 with no public AI policy, 1 with an AI-for-brainstorming-only rule, and 1 outright ban. Zero schools in the cohort permit AI drafting. Zero require disclosure. Zero describe an enforcement mechanism beyond the standard Common App fraud certification.

For context, the UC system has 10 campuses on a coordinated enforcement track. The Ivy League has 8 schools and 8 different policies. NESCAC has 12 schools and effectively 2 policies — both of them written by individual schools acting alone.

The cohort, classified

Using GradPilot's L0–L4 permission scale (L0 = silent, L4 = banned) and the supporting D (disclosure) and E (enforcement) axes, here is what every school in this cohort says about AI in admissions essays as of February 2026:

L3 = AI permitted only for brainstorming. L4 = AI banned. Read the methodology if you want the full grading rubric.

The silent ten

This is the part that needs no narration. Each of the schools below has been searched across admissions homepages, application instructions, FAQs, essay-writing guidance, and the broader institutional AI pages. None of them publish a single line of AI guidance directed at applicants.

• Williams College — No mention of AI on the first-year application page or the application supplement. Williams has robust academic AI guidance from its OIT and Writing Center, and a 2025 Ad Hoc Committee on Academic Integrity report explicitly noted that AI management in academics falls outside its scope. None of it touches admissions.
• Amherst College — The writing supplement page asks for "original, personal responses" but never names AI. Amherst has an internal generative AI task force, an AI white paper on plagiarism, and AI career-development guidance. Admissions is silent.
• Bowdoin College — Bowdoin launched an internal admissions AI pilot for data analysis in 2024 and accepted a $50M Hastings Initiative for AI and Humanity gift. Despite that, the admissions office publishes zero applicant-facing guidance. The Academic Honor Code does not mention AI.
• Middlebury College — The institutional handbook's plagiarism definition does include AI ("It makes no difference whether the source is a student, a professional, artificial intelligence…"), but it applies to enrolled students in the classroom, not applicants.
• Bates College — The "Three Tips for a Great College Application" page warns against plagiarized statements and "the work of a parent, teacher or essay-writing professional." AI is not named.
• Colby College — Notable for an enforcement-adjacent line that isn't about AI: "Colby conducts routine audits of select applications to confirm the accuracy of the information submitted." No AI-specific guidance.
• Connecticut College — "We just want you to be yourself in your content and style." Again, no mention of AI.
• Hamilton College — Their essay guidance trusts the reader's ear: "the voice of a teenager is very different from a parent's, and we've gotten pretty good at spotting the differences." That's their AI policy by implication.
• Trinity College (CT) — A 2025 institutional PDF on AI tools exists at trincoll.edu, but it is not linked from the admissions site and does not address applicants.
• UMass Amherst — Has gone out of its way to clarify that they don't use AI to read applications ("we do not use AI to evaluate or make decisions about undergraduate student applications"), but says nothing about you using AI to write them.

Ten schools. Zero policies. A few hints in honor code language. Nothing applicants can plan against.

The two that wrote something down

Tufts (L3 — brainstorming only)

Tufts is the only school in the cohort with a policy that permits AI in any form, and it is the most restrictive permissive policy you can write. From the Tufts undergraduate admissions Short Answer Questions page:

"Applicants…may use AI and other online tools for brainstorming or creative inspiration."

Followed immediately by:

"we expect that applicants will not copy or transcribe generated text or materials directly into an application."

And from the Inside Admissions blog:

"AI does not have the capacity to effectively replace an applicant's unique perspectives, experiences, and writing style."

That is the entire policy. AI may sit in the room while you brainstorm. AI may not write the sentences. There is no disclosure requirement (D0) and no described enforcement mechanism (E0). The previous Tufts statement (2024) was classified L1/D1/E2; in February 2026 it was reclassified to L3/D0/E0 after the dedicated /apply/ai-statement/ page 404'd and the language was folded into the Short Answer Questions page. The direction of travel was more restrictive, not less.

See the full Tufts AI policy entry for source URLs.

Wesleyan (L4 — banned)

Wesleyan is the only school in the cohort with a hard ban, and the policy lives in an unexpected place: the International Applicants page. From the section titled "Authenticity, AI, and the Use of Agents":

"Wesleyan recognizes there are artificial intelligence (AI) tools and resources available to candidates. Nonetheless, we expect that candidates' responses to all application questions are authentically their own and reflective of their abilities."

"Candidates may not submit as their own, any content either copied and pasted from an AI resource or authored by another individual."

"If it is determined that an applicant received inappropriate assistance with the application essays, candidacy will be compromised and/or admission rescinded."

Note the scope: "responses to all application questions." The policy applies to every applicant, not just internationals — Wesleyan just happened to write it on the international page. Wesleyan's Bachelor of Liberal Studies program separately requires that writing samples "be your own work, without benefit of AI." This is the cleanest L4/E2 (consequence-described) combination in the cohort.

See the full Wesleyan AI policy entry.

The coordination paradox

Here is what makes the NESCAC silence strange.

NESCAC is not a loose affiliation. The conference is governed by the presidents of its eleven member institutions. The presidents already have a history of coordinated policy-making in admissions: in the early 2000s, NESCAC admissions deans collaborated on a common framework for recruited-athlete admissions, ultimately producing the NESCAC Statement of Common Admission Practices. All eleven members agreed to shared guidelines about how athletic recruitment intersects with admissions decisions. The mechanism for joint policy exists. The deans talk to each other. They've done this before.

And yet on the most consequential application-integrity question of the decade — whether and how AI can be used to draft essays — there is no joint NESCAC statement, no shared language, no common rubric, and no agreement among the deans we can find publicly.

Compare to the University of California system, which runs the most coordinated multi-campus AI policy effort in the dataset: a system-wide directive that allows plagiarism checks and AI-content disqualification at all 10 campuses, with documented enforcement language at the campus level. The UC system put a policy in writing because the system office wanted one. NESCAC has the same coordinating capacity and has used it for something else.

Why the silence might exist

A few hypotheses that fit the evidence:

1. Honor-code culture is doing implicit work. Bowdoin, Hamilton, Williams, Middlebury, and Amherst all have decades-old honor codes built around the premise that submitted work is your own. Middlebury's handbook now names AI in its plagiarism definition. Several schools may believe their honor-code language already covers this and that writing a separate AI policy would be redundant or weaken the broader principle.

2. Small admissions offices may not codify what they enforce. At a school admitting 1,500 students from a pool of 10,000, the application reading process is intensely human. Readers know the voice they are looking for. Hamilton's "we've gotten pretty good at spotting the differences" is a culture, not a policy. Manual review (what we call E1) may exist in practice without being written down — making the apparent E0 classification a documentation gap, not necessarily an enforcement gap.

3. Liability avoidance. Writing down a specific AI policy creates a specific commitment. If a school publishes "we use AI detection" and an applicant is wrongly flagged, the school owns that decision. If a school publishes "AI is banned" and discovers it admitted dozens of AI-assisted essays last cycle, it owns that gap too. Silence preserves discretion. The Common App's own fraud certification — which treats submitting "the substantive content or output of an artificial intelligence platform" as application fraud — already gives every school a backstop without requiring them to write their own.

4. The cohort assumes a national policy will eventually arrive. Tufts is one of the schools whose policy tightened (L1 → L3) between cycles. Most of the schools that have published statements are doing it reactively, not preemptively. NESCAC may simply be waiting.

None of these are mutually exclusive. The most likely explanation is some combination of (1) and (2): the cohort runs on culture and discretion, not on documented policy, and culture doesn't compile to a URL.

What it means for applicants

If you are applying to NESCAC schools or the adjacent New England liberal arts schools in this cohort, treat the cohort as L3 by default:

• AI for brainstorming and reflection is reasonable. Tufts is the cohort's only stated permission level, and brainstorming is the line it draws.
• AI drafting is not safe at any school in the cohort. Wesleyan's ban is the floor; Tufts' prohibition on "copy or transcribe" is the next floor up; every other school in the cohort relies on the Common App fraud certification, which treats AI-drafted content as fraud.
• Voice is the actual enforcement mechanism. Hamilton said it out loud. Application readers at these schools read for the voice of an 18-year-old. AI-polished prose that reads like a corporate blog post fails this test before any detection tool would ever see it.
• Read each school's honor code. It is the most reliable signal in the absence of a written AI policy. Middlebury, Williams, Amherst, and Bowdoin honor codes are short, public, and worth two minutes of your time before you submit.
• Disclosure norms don't exist here. No school in this cohort asks you to disclose AI use, so there is no checkbox to check. The Common App fraud certification is the operative floor.

The broader picture

NESCAC's silence is the inverse of the Ivy League's fragmentation. Eight Ivies wrote eight different policies. Twelve NESCAC-adjacent schools wrote two — and the two were written by individual admissions offices, not by the conference. The UC system wrote one policy and applied it to 10 campuses. Three of the country's most visible school cohorts have produced three completely different responses to the same question.

If you're trying to understand where the country is on AI in admissions, the answer is "not anywhere coherent yet." Browse the full AI policies directory to see how spread out the 174 schools we track actually are. The cohort patterns — Ivy fragmentation, UC coordination, NESCAC silence — are the operative units, not the individual policies.

The silent ten in this cohort are not making a statement by not making a statement. They are betting that culture is enough. That bet has been a good one for two centuries. It is being tested for the first time this admissions cycle.

---

External sources cited:

• Tufts Short Answer Questions — primary source for Tufts L3 quote ("Applicants...may use AI and other online tools for brainstorming or creative inspiration" and "we expect that applicants will not copy or transcribe generated text or materials directly into an application").
• Tufts Inside Admissions blog — secondary Tufts quote ("AI does not have the capacity to effectively replace an applicant's unique perspectives, experiences, and writing style").
• Wesleyan University International Applicants page — primary source for the three Wesleyan L4 quotes.
• NESCAC Statement of Common Admission Practices, hosted by Colby — evidence that NESCAC has a coordinated joint-policy mechanism for admissions (used for athletics).
• New England Small College Athletic Conference (Wikipedia) — background on NESCAC governance by member-college presidents.