For educators working on the front lines of college preparation, the landscape of higher education admissions has long been fraught with visible fault lines. Advisors coaching first-generation, low-income (FGLI) students are well-versed in helping applicants navigate a gauntlet of controversial and systemic hurdles, ranging from the seismic shifts of the Supreme Court’s affirmative action rulings to the rising reliance on trauma narratives in personal statements and the shifting tides of federal and institutional diversity policies. Yet, as a new admissions cycle gathers momentum, a quiet and deeply unsettling anxiety has pushed its way to the forefront of student concerns: the arrival of artificial intelligence in the writing process.
Today, students are increasingly consumed by the fear that colleges are subtly expecting them to demonstrate an innate, sophisticated level of "AI literacy" within their admissions essays, long before they ever step foot on a university campus. This pervasive anxiety is not driven merely by the existence of the technology, but rather by a profound and systemic lack of transparency from the institutions reviewing these applications. The Common Application explicitly classifies unacknowledged AI-generated content as a form of academic fraud. However, individual colleges and universities typically refuse to disclose how their proprietary detection systems actually operate. Caught in this shroud of uncertainty, applicants find themselves frantically studying the stylistic prose patterns of modern chatbots—not to mimic them, but to desperately ensure they do not accidentally trigger a hidden institutional AI radar. The strategy has become an exercise in reverse engineering: knowing the technology intimately just to ensure your natural voice does not sound like a machine.
This fear of accidentally sounding like AI is further complicated by a fragmented landscape of institutional policies that often contradict one another. While some universities strictly forbid any algorithmic assistance, others explicitly encourage students to utilize AI as a collaborative tool "to support your original ideas," or permit basic grammar and proofreading software like Grammarly. These divergent stances fundamentally alter the calculus of competitive admissions. If select institutions actively permit or encourage AI collaboration, choosing not to use these tools could inadvertently place an applicant at a distinct competitive disadvantage. To compound the confusion, a growing number of colleges are introducing supplemental essay prompts directly addressing AI, requiring applicants to discuss the technology intelligently regardless of whether they employ it themselves.
For an affluent student with access to specialized coaching, these shifting waters are merely another nuance to manage. But imagine navigating this complex maze from an underfunded school district where no one has ever formally taught you how to interact with these tools ethically or effectively.
For years, educational researchers and equity advocates have understood that the primary metrics deciding college entry are inherently skewed toward the wealthy. Standardized test scores like the SAT and ACT exhibit a direct and stubborn correlation with household income. Similarly, critical resume-builders like Advanced Placement (AP) and International Baccalaureate (IB) courses remain vastly more accessible to students in well-funded suburban school districts than to their peers in resource-starved rural or urban systems. AI literacy has rapidly become the newest manifestation of this structural divide, yet the higher education sector has been slow to address how the digital AI divide is actively reshaping admissions outcomes.
Within the broader conversation surrounding K-12 education, generative AI has emerged as the definitive dividing line. Affluent, white-collar families, acutely aware of the paramount importance of AI proficiency in future labor markets, are aggressively lobbying their local school districts to incorporate comprehensive AI literacy into core curricula. Economically disadvantaged school districts, conversely, lack both the community advocacy and the foundational funding required to integrate sophisticated AI tools or structured training into their classrooms.
This disparity is borne out by empirical data. A comprehensive study published by the College Board, drawing on responses from more than 10,000 students, parents, and educators, revealed a stark contrast in institutional access. Private school teachers reported vastly greater access to generative AI tools than their public school counterparts, and private institutions were found to be more than twice as likely to both permit the use of generative AI and maintain formal policies governing it. Because household income and race remain deeply intertwined within the broader socioeconomic fabric of the country, experts warn that this digital divide will inevitably widen existing racial and economic inequalities. As sociologist Jerel Ezell has noted, white and affluent students are positioned not only to secure first access to cutting-edge technology, but also to gain an insurmountable upper hand on the foundational skills that will propel the future economy.
Educators witness this stark divide on a daily basis. FGLI students simply do not possess the same intuitive technological know-how or institutional scaffolding as peers growing up in wealthy communities. They generally do not have parents who enroll them in specialized robotics summer camps, nor do they benefit from high school curricula tailored around algorithmic proficiency. Instead, underprivileged students are frequently left to navigate these emerging tools entirely on their own, often without institutional oversight, ethical guardrails, or cultural guidance.
This unstructured experimentation can easily lead to the very pitfalls and missteps that quietly filter underprivileged students out of the admissions pool. Last month, for instance, a student submitted essay drafts that bore the unmistakable stylistic markers of a large language model. When confronted about the provenance of the text, the student revealed that the drafts had already been reviewed and approved by his high school guidance counselor. The immediate reaction from advisors is often one of disbelief—how could a professional educator miss such an obvious algorithmic footprint? The answer lies in systemic strain. In many underfunded public school systems, guidance counselors manage staggering caseloads exceeding 700 students, leaving them with mere minutes to spare per applicant, while other schools lack a dedicated college counselor altogether. By contrast, a counselor working within a well-resourced district is far more likely to possess the time, bandwidth, and technological support infrastructure necessary to evaluate student drafts for AI usage and teach proper citation.
This reality highlights a fundamental oversight among higher education institutions: stumbling through the independent discovery of an AI chatbot is entirely different from learning ethical utilization from a trained educator. True AI literacy extends far beyond the ability to afford a monthly ChatGPT subscription; it is rooted in cultural osmosis, regular curriculum integration, and direct, guided proficiency training.
As the current admissions architecture stands, the wildly differing AI policies among universities inevitably privilege the wealthy. This structural bias is most glaringly evident in the realm of detection systems. Most colleges maintain a policy of strict nondisclosure regarding the specific software they use to screen applicant essays, nor do they explain how they train their admissions officers to intuitively evaluate text for algorithmic signatures. Regardless of this secrecy, both automated and human policing systems are notoriously unreliable. Commercial detection software has repeatedly been shown to suffer from high error rates and documented biases against nonnative English speakers. Human detection is similarly flawed, depending entirely on the subjective impressions and baseline AI literacy of individual admissions officers.
Compounding this lack of accountability, if a student is rejected on the suspicion of unauthorized AI use, institutions are under no obligation to disclose that rationale in a rejection letter. Colleges retain the authority to quietly discard applications that simply "feel" like AI without ever being required to publish their detection rates, error margins, training data, or screening protocols.
Critics argue that if universities cannot definitively prove that their AI detection methods are equitable and accurate, they have no ethical justification to penalize applicants for suspected AI usage. Greater transparency would require institutions to clearly state whether they rely on detection programs like Turnitin, GPTZero, or Copyleaks, whether they scan for algorithmic watermarking, how frequently they discard applications based on these screens, and what data trains their internal review processes. Such disclosures would go a long way toward leveling the playing field for students who lack dedicated admissions support networks.
While some administrators worry that transparently sharing detection methods might inadvertently instruct applicants on how to evade scrutiny, the reality of the current status quo is far more damaging. As long as the mechanics of AI detection remain shrouded in mystery, the college essay ceases to be a leveling, holistic metric and instead transforms into yet another focal point for systemic inequality and bias. As artificial intelligence grows increasingly sophisticated, these disparities will only compound.
Asking colleges to bring transparency to their behind-the-scenes approach to AI is hardly an unreasonable demand. Universities already place the burden of proof on applicants, expecting them to demonstrate their readiness and literacy. It is past time for institutions of higher education to demonstrate theirs.
Finn Anderson is director of writing at Leadership Enterprise for a Diverse America and a lecturer in discipline in the Undergraduate Writing Program at Columbia University.
