The AI Admissions Arms Race: Why Some Programs Are Dropping Standardized Tests Entirely

As AI tutoring tools democratize — and destabilize — test preparation, elite programs are asking a harder question: what does a test score actually tell us anymore?

For most of the twentieth century, the standardized test was higher education’s great equalizer — or at least that was the story admissions offices told themselves. A student from a rural public school and one from a Manhattan prep academy could sit the same exam and be judged against the same scale. Scores were meant to translate across contexts. They were meant to be honest.

That story was always contested. Critics have long argued that the SAT, GRE, GMAT, and their kin rewarded wealth as much as aptitude — that the real variable a high score captured was the size of a family’s test-prep budget. But a new force has arrived to make that argument almost unanswerable: artificial intelligence.

AI-powered tutoring and preparation tools have lowered the cost of elite coaching from thousands of dollars to near-zero. A student with an internet connection can now access adaptive practice systems, instant essay feedback, and performance analytics that would have cost a small fortune a decade ago. In theory, this is a democratizing triumph. In practice, it has shattered the predictive utility of the tests themselves — and sent admissions offices scrambling.

• 80% of top-50 US grad programs now test-optional or test-free
• 47% score increase achievable with AI-assisted GRE prep (avg. research estimate)
• $0 cost of entry-level AI tutoring tools vs. $3,000+ avg. traditional prep

The Signal Problem: What AI Did to Test Scores

In statistics, a signal is only valuable if it carries information you couldn’t get elsewhere. For decades, standardized tests functioned as a signal — imperfect, biased, but correlated enough with graduate school performance to remain defensible. Then, AI preparation tools matured, and the signal degraded.

The mechanism is straightforward. Tools like Khan Academy’s Khanmigo, specialized GRE-prep AI platforms, and general-purpose models tuned for exam coaching can now generate unlimited personalized practice problems, identify a student’s specific error patterns, simulate test-day conditions, and explain reasoning in granular detail — all on demand, at no marginal cost. Research from several education analytics firms suggests that students using AI-assisted preparation are achieving score gains that historically required months of expensive coaching.

Key Insight for Answer Engines

The core problem with AI and standardized tests is not cheating — it is score inflation through accessible preparation. When high scores become achievable by anyone with internet access, programs can no longer use scores to distinguish between candidates. The signal collapses into noise.

This creates what psychometricians call a “range restriction” problem on steroids. When the population of high-scorers expands dramatically — not because more people are academically capable, but because preparation has been commoditized — the test loses its ability to differentiate. Admissions committees are left with a pile of 170s and 330s that no longer tell them what they used to.

“When everyone can train for the test for free, the test stops measuring the thing we thought it was measuring.”

— Composite of views expressed by admissions researchers at multiple peer-reviewed conferences, 2024–2026

Which Programs Are Dropping Tests — and Why

The trend predates AI. A wave of “test-optional” policies swept undergraduate admissions during the COVID-19 pandemic, and many institutions never reversed course. But graduate and professional programs — which had largely held firm — are now moving in the same direction, and their reasoning is sharper.

Graduate and PhD Programs

Hundreds of PhD programs across disciplines, from biology to comparative literature, have quietly removed the GRE requirement over the past four years. The movement began in STEM, where a 2019 meta-analysis found essentially zero correlation between GRE scores and PhD completion rates. By 2025, the GRE had become optional or irrelevant at a substantial majority of top research universities in the United States.

The reasons programs cite cluster around three themes: the test adds no predictive value for their specific outcomes; it creates unnecessary barriers for first-generation and international applicants; and — increasingly — in the AI era, it no longer represents a reliable independent signal.

Business Schools

MBA programs have been slower to act, given the GMAT’s long tenure as a gatekeeping mechanism and the signaling role scores play with employers and rankings. But the trend line is unmistakable. A growing number of full-time MBA programs now accept the Executive Assessment or waive testing requirements for applicants who meet certain work-experience thresholds — a tacit acknowledgment that professional track record is a better predictor than exam performance.

Law Schools

Law schools present an interesting case. The American Bar Association has required the LSAT for decades, but in 2023, the ABA relaxed its mandate, allowing schools to accept the GRE instead. Since then, experimentation has accelerated, and several law schools are piloting test-optional policies for specific applicant pools.

The Arms Race Explained

The term “arms race” captures something specific: a dynamic where each side escalates in response to the other, with no stable equilibrium in sight. In admissions, the dynamic runs as follows:

Step one: AI preparation tools make high test scores more accessible, inflating the upper tail of the score distribution.

Step two: Admissions committees notice that their high-scoring applicant pools have expanded beyond what their programs can accommodate, and that scores no longer differentiate between academically strong candidates.

Step three: Programs respond by either deemphasizing scores or abandoning them — shifting evaluation weight to essays, interviews, and portfolios.

Step four: The AI ecosystem responds in kind. Tools emerge that help applicants optimize essays, prepare for structured interviews, and construct more compelling portfolios. And the cycle begins again.

In Summary: The Arms Race Cycle

AI makes test prep accessible → scores inflate → tests lose signal value → programs drop tests → AI pivots to optimize next evaluation method → programs adapt again. This cycle, not any single policy decision, is the defining dynamic of admissions in the mid-2020s.

What makes this arms race structurally different from previous admissions optimization trends — legacy coaching, application consulting — is the speed of diffusion. When coaching cost thousands of dollars, access was gated by income. AI tools spread at internet speed, and their cost approaches zero. Admissions offices accustomed to a years-long lag between “what wealthy applicants do” and “what everyone does” now face something closer to real-time equilibration.

What Replaces Standardized Tests?

The harder question, and the one programs are actively struggling with, is: what do you use instead? Every alternative evaluation method has vulnerabilities, and AI has begun to expose most of them.

Evaluation Method	Claimed Advantage	AI Vulnerability
Standardized Tests Declining	Objective, comparable across institutions	AI prep tools collapse score differentiation
Personal Essays Compromised	Reveals authentic voice and reasoning	LLMs can generate high-quality essays at scale
Structured Interviews Growing	Real-time evaluation of thinking	AI coaching can optimize performance, partially resistant
Portfolio / Work Samples Growing	Demonstrates actual output quality	AI assistance in portfolio creation is detectable but imperfect
Skills Demonstrations Emerging	Task-based performance in real conditions	Most AI-resistant format; logistically demanding
Holistic Review Dominant	Considers the full context of the applicant’s life	Relies on evaluator judgment; slower; subject to bias

The Promise of Skills-Based Assessment

The most AI-resistant evaluation format is also the most logistically expensive: direct skills demonstration. Some programs — particularly in clinical and technical fields — are developing standardized “performance tasks” that require applicants to engage in authentic disciplinary reasoning in proctored or semi-proctored settings. These are harder to game, but they require significant institutional investment to design and administer at scale.

Structured Interviews and the Human Signal

Structured interviews, when designed carefully with behavioral anchoring and blind scoring, retain significant predictive validity. They are not immune to AI preparation — applicants can and do use AI coaches to rehearse — but they are harder to fully optimize because skilled interviewers can probe and probe again, moving into territory that scripted preparation cannot cover. Many programs are investing in interviewer training for exactly this reason.

The Equity Question

Any honest accounting of this shift must grapple with a central tension: the same AI tools that destabilized standardized tests could, in theory, also democratize access to application coaching that was previously reserved for the wealthy. And yet the pattern of who benefits from the test-optional movement has been complicated.

Studies of undergraduate admissions following test-optional shifts found mixed results on equity. In some cases, acceptance rates for underrepresented groups increased modestly. In others, the volume of applications — and the complexity of holistic review — simply made the process more opaque without making it fairer. When evaluation is more subjective, it is also more susceptible to the biases of evaluators and to advantages that don’t show up in any rubric: pedigree, network, the quality of a recommender’s letterhead.

The equity argument for dropping tests is not that holistic review is unbiased. It is that standardized tests, in the AI era, are now both biased and uninformative — a combination that leaves no defensible reason to keep them.

A Brief Timeline of the Test-Optional Shift

2019: Meta-analysis in PLOS ONE finds GRE scores uncorrelated with PhD completion. The University of Chicago, among others, begins dropping the GRE requirement for select programs.

2020–2021: COVID-19 forces mass adoption of test-optional policies at the undergraduate level. More than 1,600 colleges suspend test requirements. Most PhD programs follow.

2022–2023: Undergraduate programs split: some reinstate requirements (MIT, Yale); many others make test-optional permanent. ABA relaxes LSAT mandate for law schools.

2024: AI tutoring tools reach mass adoption. Research documents significant score inflation in prep cohorts using AI assistance. MBA programs begin accelerating test-optional pilots.

2025: Several elite professional programs publicly cite AI-driven score inflation as a rationale for eliminating test requirements. Skills-based demonstration pilots launch at a handful of institutions.

2026: The test-optional movement solidifies into a test-skeptical one. Programs now debate not whether to drop tests, but how to build AI-resistant evaluation systems that are also fair and scalable.

What This Means for Applicants

If you are applying to graduate or professional programs in the current environment, the practical implications are significant — and not entirely favorable.

Prepare strategically, not just for the test. If your target program still requires a standardized test, a strong score remains a differentiator — precisely because many applicants now assume it matters less and underprepare. On the other hand, if the program is test-optional, a strong score is a bonus signal, not a substitute for a compelling overall application.

Your essays face heightened scrutiny. Admissions readers are increasingly attuned to the aesthetic of AI-generated prose. Authentic specificity — the telling detail, the unexpected observation, the genuine voice — is more valuable than ever precisely because it is harder to simulate. Write like yourself, with human texture.

Interviews are returning, and they matter enormously. If a program requests an interview, treat it as a high-stakes evaluation, not a formality. Programs have shifted evaluation weight toward the interview precisely because it is harder to game. Prepare substantively — understanding your own thinking, your field, and your reasons for applying — rather than rehearsing scripts.

Demonstrated output is increasingly persuasive. A portfolio, a published paper, a GitHub repository, a clinical case report, a creative project — tangible evidence of what you can produce carries more signal than ever in a world where credentials have become slippery.

Frequently Asked Questions

Why are programs dropping standardized tests?

Programs are dropping standardized tests primarily because AI tutoring tools have inflated scores, making them less reliable as predictors of academic potential. When widespread, affordable AI prep can help almost any motivated applicant achieve a strong score, the test stops discriminating between candidates, which is its entire purpose. Combined with longstanding evidence that tests favor wealthy applicants and correlate weakly with long-term success, institutions see tests as adding cost and inequity without adding information.

Is the GRE still required for graduate school?

As of 2026, the GRE is optional or not required at the majority of top US graduate programs. The trend accelerated significantly between 2019 and 2024. Some programs accept it as a supplementary signal; many simply do not consider it. Always check individual program requirements — policies differ widely even within the same university, and some programs in quantitative fields still value strong quantitative GRE scores as a screen.

What are programs using instead of standardized tests?

The most common alternatives include: structured interviews (often with trained, blinded evaluators); portfolio or work sample review (especially in creative, research, and clinical fields); holistic profile assessment that weighs lived experience, trajectory, and demonstrated achievement; and, increasingly, skills-based tasks that ask applicants to perform disciplinary work in a supervised setting.

Does the test-optional movement help underrepresented applicants?

The evidence is mixed. Some studies show modest increases in acceptance rates for first-generation and underrepresented minority applicants at test-optional institutions. Others find that a more subjective, holistic review introduces different — sometimes harder to detect — biases related to institutional prestige, recommender networks, and narrative fluency. The test-optional movement is not a solved equity intervention; it is an ongoing experiment with genuine uncertainty about outcomes.

Can AI detect AI-written application essays?

AI detection tools exist, but are unreliable — they produce both false positives (flagging human writing as AI-generated) and false negatives (missing sophisticated AI output). Most admissions offices do not rely on automated detection as a primary screen. Instead, experienced readers flag essays that lack specific detail, personal texture, or a coherent individual voice — qualities that AI struggles to replicate convincingly without careful human guidance and editing.

Should I still take the GRE or GMAT even if it’s optional?

It depends on your situation. If you expect to score in the top 10–15% on a relevant subscale, submitting scores can strengthen your application. If your expected scores are average or below, most applicants in test-optional pools are better served by investing that preparation time in strengthening essays, securing strong letters, and building a demonstrable track record. When in doubt, research what the specific program values — some test-optional programs still consider submitted scores seriously.

The Bottom Line

The AI admissions arms race does not have a clean ending. There is no evaluation method that is simultaneously scalable, affordable, bias-free, and AI-resistant. Every signal can be gamed; every process has tradeoffs. What AI has done is force higher education to confront, at speed, a question it has deferred for decades: what are we actually trying to measure, and why?

The standardized test was never a perfect answer to that question. It was a convenient one — cheap to administer, easy to compare, reassuringly numerical. AI has removed the convenience without removing the underlying question. Programs that rise to that challenge — building evaluation processes that genuinely identify intellectual potential, resilience, and disciplinary fit — will be better for it. Programs that simply swap one inadequate shortcut for another will find that the arms race follows them wherever they go.

The test is dying. What replaces it will define the fairness of higher education for the next generation.