The conversation in most families about AI and college admissions runs in one direction: should my student use ChatGPT on the essay? The Common App says no, the schools say no, the detection tools are unreliable, and the question dominates parent discussions.

The conversation that is not happening runs in the other direction. Selective US universities are already using AI to read applications, score essays, summarize files, interview applicants, and triage tens of thousands of submissions before a human ever opens them. This is documented, recent, and being expanded for the 2026 cycle.

If your student is applying to a selective school next year, this changes how the application should be built.

What is actually happening

The dominant CRM in selective admissions is already running AI. Slate, made by Technolutions, is the back-end admissions system used by more than 2,100 colleges, universities, and secondary schools. In 2024, Slate launched Reader AI, a tool that pre-reads application documents and summarizes them for human reviewers. Per Slate’s own marketing, the tool “can quickly analyze and identify pertinent details in letters of recommendation, college essays, or any other document you provide.” Schools that use Slate can enable Reader AI for any document type, including essays, recommendations, transcripts, and research uploads.

UNC has been auto-scoring essays since 2019. Per Daily Tar Heel reporting, UNC’s Office of Undergraduate Admissions uses an automated essay-scoring system that evaluates writing quality including vocabulary, sentence structure, grammar, and length. The auto-generated score appears on the Slate Reader Dashboard alongside the application. Human readers see it before they make their evaluation.

Virginia Tech debuted an AI essay reader for the current cycle. Per Virginia Tech News, the AI is being used “to confirm the human reader essay scores, not make any admissions decisions.” The university stated purpose is to deliver decisions one month earlier than the previous schedule.

Caltech ran a new AI-powered interview system this past cycle. Per Caltech admissions dean Ashley Pallie, the tool, called VIVA and developed by a company called InitialView, was used to screen roughly 10 percent of early applicants who submitted research projects. Students upload their research and an AI voice interviews them on video, asking questions about the work, akin to a brief dissertation defense. Faculty and admissions officers then review the video. Pallie has stated that Caltech plans to expand its AI use for 2026.

Per an Inside Higher Ed survey, roughly half of US admissions offices already use AI in some part of the application review process. The most common uses are transcript review and recommendation-letter summarization. Essay reading is less common but growing.

What the policies actually say

The official institutional language is consistent. Every school using AI tells the same story: AI assists human reviewers but does not make final decisions. Slate’s documentation explicitly states that Reader AI evaluates a single application at a time and does not compare or rank applications. UNC, Virginia Tech, and Caltech all stress that human readers are still the decision-makers.

This is true. It is also incomplete. The AI does not decide who is admitted. It does decide what the human reader sees first, how the file is summarized, and how quickly the file moves through the queue.

In a process where the average human reading time is approximately eight minutes per application, the AI summary is increasingly what the eight minutes is spent on. The human reader is not starting from your student’s essay. They are starting from the AI’s pre-read of the essay.

What this means for your student

Three practical implications.

First, applications now need to be readable by both AI summarizers and human readers. AI summarizers favor clear structure, named projects, specific facts, and identifiable claims. They struggle to summarize narrative essays that depend on subtext or implication. The student whose application includes a clear specific story, a named focus, and concrete accomplishments is more likely to come through the AI pre-read with the right summary attached. The student whose application is vague or full of generic accomplishments is more likely to come through with a flat or inaccurate summary.

Second, research projects and other student-generated work now face authenticity verification. Caltech’s VIVA is the first published example, but it will not be the last. Students who submit research should expect that the depth of their understanding can be tested, either by AI interview or by human follow-up. Submitting work the student cannot intellectually claim is now actively risky in a way it was not three years ago.

Third, the “eight minutes” of human review has not gotten longer. It has gotten more curated. The human reader still spends the same amount of time on your student’s file, but the AI is shaping which parts of the file get those minutes. Clarity is no longer just helpful. It is structural.

The conversation about AI in admissions has been pointed in the wrong direction. The question your student should be asking is not whether to use AI on the essay. It is whether the essay, the activity list, and the research uploads will read well to the AI that gets to them first.