Let’s get this out of the way. Artificial intelligence is not deciding who gets into college. Not yet, anyway.
To be sure, a growing number of college admissions offices have started incorporating AI agents into their operations, but conversations with several companies selling those products suggest that none of the schools have handed the keys over to the machines completely. If anything, admissions offices are underutilizing AI in the admissions process out of a concern for keeping control over who shows up on their campus and how much they will pay. That probably will not change in the next few years, but the need to increase efficiency and reduce expenses will inevitably drive admissions offices to expand the use of AI at every stage of their work. Technology companies are lining up to meet their needs.
It is difficult to know who is taking them up on which offerings. A cursory review of college websites revealed that a growing number direct students on how they can use AI in their applications, with the guidance ranging from cautionary guardrails at Northwestern to outright bans at Brown. Few, if any, tell students how their admissions office is using the technology. That makes it difficult for students and families to know what, if anything, they need to do to meet the demands of an AI admissions office.
I spoke to several companies selling AI products to colleges to get a better sense of what they currently have on offer for admissions offices. Most of the products cover three areas: communication and counseling, recruitment and yield, and admissions and decisions. In general, admissions offices are currently using AI to handle important but monotonous tasks that cost admissions offices not only money but time that could be spent in direct human-to-human contact. “Colleges are being asked to do a lot more with fewer resources,” CollegeVine’s Emily Smith told me. Companies like hers and others I spoke to develop AI products so “humans are freed up to do the stuff that is better suited for human work,” like reading personal essays and making the ultimate decision on who to admit or reject.
Nor are they using AI in particularly innovative ways, Harneet Singh of Rabbit AI and TechCurators suggested to me. When I asked him if AI could be used to accomplish something like the analytics revolution in sports, where novel data and new modes of analysis have transformed teams’ rosters and strategies, he said that was certainly possible, but not really happening. His companies are working on better assessments and AI interviews that will be able to determine a whole different set of skills than most exams currently do. The tests colleges use now can “check for aptitude or check for communication [skills], but none of them check for [applicants’] thinking, for their agility, for their ability to learn.” He thinks that is where AI is going, but we are not there just yet.
While colleges are being conservative with AI in admissions, how they’re using it right now provides a glimpse of ways that the quickly evolving technology could take over much more of the admissions process.
Communication and Counseling
Applicants are most likely to notice AI’s incorporation into admissions when they hear from or talk to colleges. Communications is where many colleges first slip their toe into the AI waters. Several companies provide AI-powered chatbots for admissions websites and use AI to craft personalized emails and texts to students. CollegeVine has an AI agent that can answer calls from students around the clock or even call applicants if they reach out. When one of them named Maple called me, it sounded like a young woman with just the right amount of vocal fry to let me know that she was chill and ready to help. Smith shared that some students feel more comfortable talking to AI agents about sensitive issues, like a bad grade in a chemistry class, than they do discussing it with their own school counselor.
AI will likely play an increasingly large role in college counseling and could provide valuable assistance, especially to students at schools where the counselors don’t have the bandwidth to help hundreds of students build college lists or figure out how financial aid works. The Common Application, which allows students to apply to about 1,000 colleges using a single form, recently announced that it is partnering with the college access organization College Possible to integrate an AI advisor into the application platform.
Recruitment and Yield
Communications and counseling play an essential role for admissions offices in the all-important recruiting process. About 2.5 million freshmen start college every year; another million students transfer to a new institution. Colleges spend large sums of money to attract some of them to apply. Recruiting includes direct marketing through mail (yup, they’re still sending flyers to Gen Z), email, and texts and advertising in airports, billboards, and social media, but it also includes campus tours, webpages, college fairs, and human recruiters. Perhaps the most challenging aspect of recruitment is determining which students to pursue, when, and how.
The approach for a lot of colleges is to license a huge list of student names and data in order to blast out the same email to each of them on a preset schedule over the course of a year or more. Niche CEO Luke Skurman told me that regional private schools with 500 freshmen “are already buying, in many cases, two to three hundred thousand names at the top of their [recruiting] funnel. They can’t just go to four or five hundred thousand; it just doesn’t make sense.” What Niche, which serves both students with college search tools and colleges with recruiting tools, essentially does, Skurman believes, is promote “better matchmaking.”
AI helps in matchmaking not only by analyzing the recruitment process as it’s happening, so it can track in real time which students are indicating interest by opening emails and texts, visiting college websites, and signing up for recruiting events, but also by reacting to that analysis and crafting a personalized recruiting strategy that is more targeted and engaging and less scattershot. Smith explained that their AI recruiting software improves on the “stale, generalized content” colleges typically send by using a range of data on each student to craft individualized messages for students.
So, instead of getting the same email hundreds of thousands of students are getting on more or less the same schedule across the country, an AI agent can send a student a text at the time of day they’ve been receptive to messages in the past and make it about the chemistry program, since it knows that they’ve been looking at that department’s page. AI can also help keep students on track after they have been admitted, nudging them along the path to enrollment and even providing personalized guidance to students between enrollment and arrival on campus. These personalized recruiting approaches can also be used to convince admitted students to enroll, which colleges refer to as their yield strategy. They could also spell the end of panic-inducing form emails pressuring students to submit paperwork and pay fees that have already been taken care of or are not needed for weeks or even months.
The ultimate version of AI matchmaking might be Niche’s direct admissions program, which allows an AI agent to determine if a student who has created an account with the company meets the qualifications a college has shared to make an offer of admission to them before they even fill out an application. Other companies, including the Common Application, and several states are currently experimenting with direct admissions, to mixed success. The colleges that participate in direct admissions programs tend to have high acceptance rates, so students should not expect to see offers coming in from Georgetown or Georgia Tech.
All of this personalization during the recruiting process raises questions about potential privacy issues. Colleges have been tracking student visits to their websites and recording open rates on emails for many years now, all as part of a process designed to measure what’s called “demonstrated interest” in the enrollment industry. AI has the potential for significantly raising the stakes on demonstrated interest by making it much easier and cheaper to track students across the internet. I asked Brian Kim if we should be concerned about colleges’ AI agents looking at their social media posts and search history to create a profile and determine if an applicant is a good match for a school. Kim is well-placed to answer this question, as a researcher with deep experience in higher education admissions and the founder of Open Augments, a company that helps organizations integrate AI into their work. He pointed out that it might be possible to purchase and integrate search history data on a potential applicant, but it would also be too costly and complex for most institutions. But as for a school using an agent “to look up an [applicant’s] social media presence to see if they have said anything about a college,” he said, “that’s really straightforward and simple.” This kind of information could play a powerful role in determining whether offering an admissions spot to an applicant is likely to pay off with an enrollment. That information would be so useful and easy to get that he would assume some institutions are already using AI this way or will in the future.
Applications and Decisions
The more successful an admissions office is during the recruitment stage, the more work they create during the application evaluation stage. In 2024, the Common Application processed over 10 million applications, which was more than twice the number in 2017. This past year, both Northeastern and NYU received over 100,000 applications. State universities and other large private institutions need to deal with tens of thousands of applications every year. In order to handle this volume, many of them rely on seasonal readers who need to be trained and constantly calibrated during the process to make sure their judgments are in line with their peers. Readers are also provided with data profiles of each applicant that allow them to quickly compare the file in front of them to other people in the student’s school and other applicants. A seasonal reader at midwestern public university told me that between the profiles and the training he felt as if his reading was essentially automated.
Data analysis is nothing new in college admissions, and many of the AI products offered to colleges are intended mainly to increase an office’s efficiency in handling documents. Jeff Selingo, author of Dream School, described transcript handling as one of the most time-consuming elements he observed in admissions offices. That’s because there are almost as many transcript formats as there are high schools. Before an admissions officer can understand what an applicant’s transcript is telling them, they have to figure out first what they are even looking at. Several companies have developed AI products that can not only extract class names and grades out of transcripts but also provide context on what those grades might mean at a particular school. For instance, if a student has “only” two AP courses, an AI report on the transcript can look at information on the school to determine that students there are limited to two AP courses.
AI tools can go further than that. Vantage Labs provides a tool to admissions offices that looks for patterns in transcripts, such as a record of improving grades, which the company’s vice president of global operations, Paul Edelblut, said they have found can be as strong or even stronger a signal of an applicant’s potential as the GPA itself. For decades, Vantage has been developing machine-reading technologies that could evaluate and score admissions exam essays. That experience positioned them well to create tools that could analyze letters of recommendation or even a complete applicant file.
For some institutions, that analysis results in extracted data points; for others, it might also include a summary report, and, at a few of them, Vantage’s AI provides recommendations on how an admissions office should proceed with an applicant, based not just on the file but on an institution’s priorities. Edelblut made it clear that their AI is not rendering admissions decisions. “Even in the most robust system, we are typically categorizing the applications into something like a green, yellow, and red light grouping.” The green ones, he continued, are the ones that the AI is saying, “Hey, a human should review this file soon so you can reach out to that applicant.” But he also noted that one of the strengths of an AI evaluation is that it can achieve a level of consistency and cool consideration that is much harder to accomplish with a large staff of emotional human beings. While it’s important to constantly audit AI systems for biases and errors, the same goes for people. It is imaginable that increased use of AI could make admissions fairer and more transparent.
What AI Means for Applicants
Everyone I spoke to echoed Edelblut’s insistence that it is human beings who are making the admissions decisions, and they do not think that is likely to change. Most of the uptake for AI has been essentially duplicative of human work and directed at drudgery, like evaluating transcripts or determining which credits a transfer student can carry over to an institution. Handing this kind of labor over to AI, Kim believes, can “extend and enhance” the work humans do and the values institutions hold, rather than replace both of them.
I hope he is right, but I am not so sure. The intense pressure on admissions offices to maintain or grow enrollment numbers and revenue at the same time their budgets are being tightened and their staff are dealing with serious burnout creates a scenario where it is not hard to imagine humans playing a reduced role in the admissions process. The most selective colleges almost certainly won’t cede control over their admissions decisions to machines, but for the vast majority of colleges, which are especially stressed and admit more applicants than they deny, I suspect the allure of AI will be hard to resist.
From an applicant’s perspective, the most significant potential change in the admissions process might be in the measurement of demonstrated interest. Families in the know have been aware for years now that colleges are tracking applicants’ engagement as a way to determine whether it is worth making an offer of admission to them. AI could hypercharge that tracking, and so students might want to think more carefully about how they present themselves online. For years, counselors have been warning students that admissions offices are checking out their social media accounts, and for years that’s been little more than a way to create unnecessary anxiety. Admissions officers simply don’t have the time to look at every applicant’s Instagram account. With AI, however, they might now be able to hand that job off, looking less for offensive content and more for signs that an applicant is interested in their school.
When I asked Selingo what he thinks families and students should do in response to increased use of AI by admissions offices, his advice was to do nothing different for now. Things have not changed enough for families yet. He doesn’t think that it’s significant that a machine is simplifying transcripts; if anything it’s good for students because it means readers have more time and energy to look for other interesting elements in a student’s file. That’s why, when I asked him how transparent colleges should be about their own use of AI tools in admissions, he believed that it does not matter if they do not tell them about things like document processing . “If AI starts being used in decisions, however,” he said, “they need to let students know.” Maybe we should give Maple a call and ask her.
