Do Academic Credentials Still Predict Professional Competence After Generative AI?

Imagine a world where a university degree is less a certificate of what a student has learned and more a signal of their ability to endure a difficult process. For decades, economists and educators have viewed academic credentials through this lens: the idea that earning a degree proves a student has the discipline and intelligence to complete a challenging task, even if the specific work they submitted was not entirely their own. This system relies on a simple assumption: that the work a student turns in is a genuine reflection of their own mind. But in late 2022, a new technology arrived that threatened to break this assumption. Generative artificial intelligence, capable of writing essays, solving problems, and drafting complex arguments, suddenly made it possible for students to produce high-quality work with a fraction of the effort. The immediate fear was that this technology would hollow out the value of degrees. If anyone could produce the work, the signal would become meaningless, and the link between a diploma and actual professional skill would snap.

This question is not just about academic dishonesty; it is about whether the entire system of professional certification still works. When a law school graduate sits for a bar exam, or a nursing graduate takes a licensing test, they are taking a high-stakes, supervised test that cannot be outsourced to a computer. These exams are the final checkpoint, the moment where the promise of a degree is tested against reality. If the coursework leading up to those exams was compromised by artificial intelligence, the degree might no longer predict who can actually do the job. The concern is that the technology has lowered the cost of producing the signal so much that it no longer separates the capable from the incapable, leaving employers and the public with a credential that no longer means what it used to.

A team of researchers set out to test this fear with hard data, looking at whether the arrival of generative artificial intelligence actually broke the link between academic credentials and professional competence. They focused on two distinct professions: law and nursing. These fields were chosen because they sit at opposite ends of the spectrum regarding how much their daily work relies on language and text. Law is a field heavily exposed to language-based tasks, making it the most likely candidate for disruption by text-generating tools. Nursing, while requiring critical thinking, involves a significant amount of hands-on, physical, and interpersonal work that is much harder for a computer to simulate. By comparing these two fields, the researchers could see if the erosion of credentials happened more in the field that was most exposed to the technology.

The researchers built a massive dataset tracking thousands of law schools and nursing programs across the United States. They looked at the academic records of students when they first entered these programs—specifically their standardized test scores and college grades—and then tracked how those same cohorts performed on their professional licensing exams three years later. This created a clear timeline: the researchers could see if the students who entered with strong academic records still became the strongest professionals after their education, and whether this relationship changed after artificial intelligence became widely available. They examined data from 2014 through 2025, covering nine years before the technology arrived and three years after.

The results were surprisingly clear and contrary to the prevailing anxiety. The researchers found that the connection between entering academic ability and professional licensing success did not weaken after the arrival of generative artificial intelligence. In fact, the relationship remained just as strong, and in some statistical measures, even slightly stronger, in the years following the technology's release. The data showed that students who entered law school with higher test scores continued to pass the bar exam at higher rates, regardless of whether they had access to artificial intelligence during their studies. The same pattern held true for nursing programs. The researchers calculated that if artificial intelligence had caused a significant drop in the predictive power of these degrees, it would have been easily visible in the data. Instead, the evidence ruled out any drop larger than about twenty percent, and the most likely scenario is that the drop was much smaller or non-existent.

One of the most striking findings was that the only major break in the data did not happen when artificial intelligence arrived. The researchers noticed a shift in the data around 2022, but this shift occurred in a cohort of students who had already finished most of their coursework before the technology was publicly released. This suggests that the disruption was caused by the pandemic and the chaotic shift to remote learning, not by artificial intelligence. The students who actually used the technology during their studies did not show a decline in their ability to pass their final, supervised exams. The data indicates that while the technology might have changed how students completed their daily homework, it did not destroy the underlying ability of the degree to predict who would succeed in the real world.

The study also looked at whether the technology helped weaker students catch up, effectively "pooling" everyone together so that the distinction between high and low performers disappeared. This was a specific fear: that artificial intelligence would allow struggling students to produce work that looked just as good as the work of top students, flattening the curve of performance. The researchers found no evidence of this. The distribution of scores did not compress in the way the theory predicted. The top performers remained distinct from the lower performers, and the gap between them did not vanish. The technology did not act as a great equalizer that erased the differences in student ability; the ranking of students by their eventual professional competence remained intact.

What does this mean for the value of a degree in the age of artificial intelligence? The findings suggest that the system of professional credentialing is more robust than many feared. The final, supervised exams that determine who gets to practice law or medicine acted as a reliable filter, catching any loss of competence that might have occurred during the earlier, unsupervised coursework. The degree still signals that a person has the capability to pass the test that matters most. However, this does not mean the system is perfect or that the integrity of daily homework is safe. The researchers point out that the risk has simply moved upstream. The danger is no longer that the final exam will be compromised, but that the learning process itself—the daily assignments and the development of skills—may have been altered in ways that the final exam cannot see. The degree still predicts success on the test, but it may no longer guarantee that the student learned everything they were supposed to learn along the way.

The study concludes that the arrival of generative artificial intelligence has not yet destroyed the signal of academic credentials. The link between what a student knows when they start and what they can do when they finish remains strong. The technology has not made the degree meaningless, nor has it blurred the line between the capable and the incapable. Instead, the system has absorbed the shock, with the final, high-stakes examinations continuing to serve as a trustworthy measure of professional readiness. The real challenge for educators and institutions is not to protect the final exam, which has proven resilient, but to ensure that the learning happening in the years before that exam is still genuine and effective. The degree still works, but the journey it certifies may have changed in ways we are only beginning to understand.