The Last Number Standing

While the LSAT score is being pushed up from below by accommodations and cheating, something else is eating the rest of the application from above.

In 2025 and 2026, admissions officers started noticing something strange about the essays coming in.¹ The personal statements got better. Not the way a strong applicant pool gets better, where some are powerful and some are flat. These all read like the same person wrote them. Well-structured, clean, appropriate, and almost without personality. Good the way a press release is good.

They were, of course, largely AI-written. Not all of them, and probably not even most. But enough to make the whole signal unreliable, because you can’t tell which ones.

Admissions offices tried detection tools. They didn’t work reliably. They flagged real human writing as AI, and they fell behind every new model the moment it updated. An applicant who ran a draft through ChatGPT in January might get flagged, while the same draft run through a newer version in March might not. The detection can’t be consistent, because it’s always chasing the thing it’s trying to catch.

Certification systems were tried too. Sign a statement that this is your own work. It works the way honor systems work when the payoff is large. The people who were going to cheat sign and keep going, and everyone else feels a little worse about them. Nothing changes.

Supervised in-person writing samples work in theory, but scale kills them in practice. A law school that processes 3,000 applications in a cycle can’t bring every applicant in for a timed writing test. You could require it as part of an interview, but you only interview a fraction of the pool. And you’d have to make the writing sample the actual admission decision, not just a check, or sophisticated applicants would use one approach for the supervised test and a different one for everything else.

So the tools fail, and institutions stop relying on tools that fail at scale.

Watch what happens next. It follows the logic of the situation exactly.

First the personal statement carries less weight, because admissions officers can’t trust it. Then the diversity statements get the same treatment. The “why our school” essays go next; they were the most gameable part of the application anyway, and now AI can generate a plausible school-specific paragraph from a website in about 40 seconds. Then the addenda, the character and fitness explanations, the resume summaries. Each judgment-based tool that once let a real person show through becomes noise at scale, gets quietly downweighted, then quietly dropped.

What’s left is 2 numbers, the LSAT and the GPA.

The GPA has a problem of its own.

For the 2024 entering class, the majority of schools’ median GPAs rose. The aggregate median GPA across ABA-accredited schools was up about 0.02, with median LSATs up about 0.09 in the same period. 40 of 196 schools raised both medians in the same cycle.² Grade inflation at the undergraduate level has been running for decades, and law school admissions data just shows the downstream effect.

The real problem with GPA as a lever is that it’s fixed. By the time you apply to law school, your undergraduate record is done. You can’t retake organic chemistry or replace the C you got sophomore year. The number you submit is the one you built over 4 years, when you were 18 to 22 and didn’t necessarily know it would matter this much.

So the GPA is inflating, it can’t be changed, and admissions offices still weight it heavily because it’s 1 of only 2 things left.

The other number is the LSAT, and the LSAT you can change. You can prep and retake. The highest-score rule means every attempt is worth taking, and you get 5 attempts in a 5-year window, 7 in a lifetime.³ The score works the same way at every school and is calculated the same way for every applicant. As of mid-2026, admissions offices still mostly trust it. It comes from a testing organization with standardization built in, the prep curve is well understood, and next to the essay pile, it feels like it means something.

So the LSAT absorbs the weight of the entire rest of the application.

Consider what that means in 2020 versus late in the decade.

In 2020, the LSAT was the most important part of the application file. It drove medians, which drove rankings, which drove scholarships, which drove enrollment. Everyone understood the essay mattered on the margins, the GPA was the second number, and recommendations and resume added texture. The LSAT was important, but it wasn’t everything.

In the stress test, by the late 2020s, with AI degrading the qualitative inputs, it essentially is the file, because everything else eroded.

That shift is happening at the exact moment the score itself is being inflated by forces that have nothing to do with reasoning ability. Accommodations up 2.5 times in 5 years. Online testing ended because cheating got serious enough to require it. First-time takers now a minority, with the repeat-taker surge built into the median the school is chasing.

So the score absorbing all the weight is also the score losing its reliability as a predictor. That’s the trap.

A real score, built through preparation that produces repeatable, explainable results, still signals something. A school, an employer, a bar examiner can trust it because the skill behind it holds up on the next test too. A hollow score, one that came from extra time the applicant didn’t functionally need, or from a proxy, or from a 7th retake that finally produced a number the baseline skill doesn’t support, shows up in first-year GPA, in bar passage, and eventually in the classroom crack the next chapter is about.

The score is the last number standing. Whether it still means anything depends on how you built it.

Watch: how your application actually gets read now

Law Schools Are Quietly Using AI—And It’s Changing Admissions Forever

This chapter is from Anyone Can Beat the Law School Collapse by Steve Schwartz, LSAT Coach and Founder of LSAT Unplugged. I've been coaching the LSAT since 2005.

Published July 28, 2026.