AI Is Reading Your Students’ Essays. Is Anyone Preparing Them?
Virginia Tech now runs an AI model alongside a human reader on every short-answer essay in its applicant pool. When the two disagree by more than two points on a 12-point scale, a second human reviewer steps in. The University of California system screens personal insight questions for AI-generated content, with disqualification as a stated consequence. Caltech has gone further still, stating outright that admission can be rescinded for violating its AI policy.
This is not a hypothetical future scenario for admissions offices to plan around. It is the current admissions cycle. And it raises a question most counseling offices haven’t fully reckoned with yet: if institutions are actively building infrastructure to detect a specific failure mode in student writing, who is teaching students to avoid it?
The policy vacuum is the real story
A recent Kaplan survey of admissions officers found that 68% of institutions still have no official policy on generative AI in essays, even as detection tooling accelerates on the other side of the desk. That gap, rapid enforcement infrastructure paired with minimal public guidance, puts students in a genuinely difficult position. They are being evaluated against a standard that hasn’t been clearly articulated to them.
Districts and schools are largely left to infer the standard from public statements and enforcement actions rather than any consistent, sector-wide guidance. NACAC’s updated ethics framework offers a useful anchor: the meaningful distinction isn’t AI use versus non-use, it’s whether AI generated the substance of the essay or assisted with brainstorming and editing around a student’s own thinking. That’s a workable line institutionally, but it requires deliberate instruction. It is not intuitive to a 17-year-old who has grown up with these tools as ambient infrastructure.
What detection tools are actually optimized to catch
The mechanism matters here. AI detection isn’t primarily a grammar or vocabulary check. It’s tuned to notice the absence of a specific signal: evidence of a mind working through something in real time, as opposed to arriving cleanly at a pre-formed conclusion. This is a metacognitive marker, not a stylistic one, and it points toward a curriculum gap rather than a policy gap alone.
Most essay instruction, understandably, focuses on structure: hook, narrative arc, resolution. Very little instruction explicitly teaches students to show their thinking process on the page, the confusion, the failed attempt, the specific moment something shifted. That skill is precisely what AI-generated text struggles to convincingly fabricate, and precisely what trained readers, human and algorithmic, are now calibrated to look for.
The readiness implication
This connects to a broader pattern we track closely: students are frequently strong at producing polished output and comparatively underdeveloped at articulating their own reasoning process. The AI essay landscape has simply made that gap visible and consequential in a new way. It is, in effect, a stress test that’s currently running on every applicant pool in the country, whether institutions have opted into it deliberately or not.
What this means for your building
Counseling and writing curricula that explicitly teach metacognitive reflection, not just narrative structure, give students a durable advantage that has nothing to do with gaming a detection algorithm and everything to do with genuine self-understanding translating onto the page. This is a curriculum investment with a shelf life well beyond the current admissions cycle, since the underlying skill (clear articulation of one’s own thinking) matters in college coursework, job interviews, and well past either.
Where we come in
Our work with partner institutions builds exactly this kind of instruction directly into existing college readiness programming, giving students structured practice in the specific reflective skill admissions offices are now, deliberately or not, screening for. If your counseling office is navigating this gap without a clear curriculum response yet, we’d welcome that conversation. Learn more about partnering with us.
Colleges have always wanted to understand how an applicant actually thinks. The tools now reading for it have simply gotten harder to fool with anything less than the real thing.