For decades, the college admissions process has been scrutinized as a barometer of American social mobility. From the influence of legacy status and athletic recruitment to the disparities in SAT preparation and extracurricular access, the path to elite higher education has long been paved with the advantages of wealth. Now, as the admissions landscape shifts into the age of generative artificial intelligence (AI), a new, silent barrier has emerged: AI literacy.
For first-generation, low-income (FGLI) college applicants, the stress of the admissions cycle—traditionally defined by anxieties over affirmative action rulings, trauma-centered narratives, and shifting diversity, equity, and inclusion (DEI) policies—has been eclipsed by a pervasive fear of the machine. Students are no longer just asking, "Will I get in?" They are asking, "Will the algorithm flag me as a fraud?"
The New Digital Divide in Admissions
The anxiety stems from a profound lack of transparency. While the Common App explicitly labels unauthorized AI-generated content as "fraud," individual colleges remain notoriously cagey about their detection methodologies. This "black box" approach forces students into a high-stakes game of cat-and-mouse. Applicants are attempting to decode the prose style of Large Language Models (LLMs) like ChatGPT, Claude, and Gemini, not necessarily to use them for cheating, but to ensure their own voices don’t inadvertently trigger an automated rejection.
This "fear of sounding like AI" is paradoxically compounded by the varying policies across institutions. Some universities encourage students to leverage AI to "support original ideas," while others permit the use of basic proofreading tools like Grammarly. In this fragmented landscape, the decision to use or avoid AI becomes a strategic gamble. If an applicant ignores these tools, they may be at a competitive disadvantage; if they use them, they risk being labeled a plagiarist by software that remains shrouded in mystery.
Chronology of a Shifting Landscape
The integration of AI into the application process has evolved rapidly over the past two years:
- Pre-2023: Admissions essays were primarily viewed as a test of personal narrative and grammatical competence. The "human" element was assumed, and tools were limited to basic spellcheckers.
- 2023–2024: The public release of advanced generative AI models forced an immediate, if reactionary, response. Many institutions updated their academic integrity policies, often relying on rudimentary AI-detection software that was quickly criticized for its high rate of false positives.
- 2025: A "gray zone" emerged. Colleges began introducing supplemental prompts specifically asking students to discuss their views on or experiences with AI. This marked a pivot: AI was no longer just a potential tool for cheating; it was now a subject of academic inquiry.
- Late 2025 to Present: The divide solidified. Well-resourced school districts began integrating AI literacy into their curricula, while underfunded districts struggled to provide even basic internet access. This has created a two-tiered system where some students are trained in ethical AI usage, while others are left to navigate the technology in a vacuum.
The Data: A Tale of Two Schools
The inequity of this digital divide is supported by stark data. A December 2025 College Board study involving over 10,000 participants highlighted a significant disparity in access. Private school teachers reported vastly higher access to generative AI tools compared to their public school counterparts. Furthermore, the study noted that private institutions were more than twice as likely to have formal policies governing the ethical use of GenAI.
This is not merely a matter of who has a paid subscription to a chatbot. It is a matter of institutional infrastructure. In well-funded, affluent districts, students are guided by college counselors who have the bandwidth to mentor them on the ethical boundaries of AI assistance. Conversely, in underfunded districts, counselors are often responsible for hundreds of students, leaving them with zero capacity to supervise how a student uses AI in their personal statement.
The consequences of this are real. I have personally worked with students who submitted essays that were functionally indistinguishable from AI output, only to discover that their overworked school counselors—lacking the time or training to flag the issue—had greenlit the drafts. These students are being set up for failure, not through malice, but through a systemic lack of guidance.
Implications for Social Mobility
The implications for racial and economic equality are profound. Sociologist Jerel Ezell has noted that the current trajectory risks formalizing an "AI caste system." When affluent students receive "first dibs" on both the technology and the nuanced training required to use it, they are effectively gaining a head start on the skills that will define the future global economy.
Furthermore, the "policing" of these essays is inherently biased. Research has consistently shown that AI-detection software is not only inaccurate but also disproportionately flags the work of non-native English speakers. Because these tools look for a certain "predictability" in prose, the more diverse linguistic patterns of multilingual students are frequently misidentified as "robotic" or "generated." When these software tools are coupled with the subjective, often uncalibrated "intuitions" of admissions officers, the result is a system that penalizes the very populations it claims to serve.
The Case for Radical Transparency
Colleges often hide behind the defense that disclosing their detection methods would teach students how to "game the system." This argument is fundamentally flawed. By maintaining silence, colleges are allowing the admissions process to function as an opaque, high-stakes lottery where the criteria for rejection are hidden from the applicant.
If universities are to maintain the "holistic" promise of their admissions processes, they must move toward radical transparency. This includes:
- Public Disclosure of Detection Tools: Colleges should clearly state whether they use software like Turnitin, GPTZero, or Copyleaks, and if they monitor for specific AI watermarks.
- Data Transparency: Institutions should publish aggregated data regarding how often applications are flagged, the rate of false positives, and the training protocols provided to human admissions officers.
- Equitable Policy Formulation: Rather than blanket prohibitions that disproportionately harm under-resourced students, colleges should provide clear, accessible guidelines on the ethical use of AI that do not require an expensive private counselor to interpret.
Conclusion: A Call for Institutional Literacy
Learning to use AI is not an innate skill; it is a learned behavior, shaped by cultural osmosis, regular integration, and direct instruction. When a college demands that a student prove their "AI literacy," they are essentially asking for a performance of privilege.
If colleges are going to continue using the admissions essay as a cornerstone of their evaluation process, they have a moral obligation to ensure that the process is not merely a reflection of who had access to the most sophisticated digital tutors. As it stands, the current system is not just failing to be meritocratic; it is actively entrenching the digital divide.
For years, we have acknowledged that standardized tests and AP course loads were tilted toward the wealthy. We must now recognize that AI literacy is the new, more insidious iteration of this phenomenon. It is time for colleges to stop demanding that students prove their AI literacy and start proving their own. They must demonstrate that their admissions processes are based on transparent, equitable, and evidence-based standards—not on the hidden biases of an uncalibrated machine.
Finn Anderson is the director of writing at Leadership Enterprise for a Diverse America (LEDA) and a lecturer in the Undergraduate Writing Program at Columbia University.
