Harvard has finally said publicly what most people in higher education have known for years: the A has become too common to mean what it once meant.
Faculty at Harvard College voted to limit flat A grades to 20 percent of enrollment, plus four additional students, beginning in fall 2027. Harvard also plans to use an internal percentile ranking for honors and prizes. It is a serious attempt to fix a serious problem, but it shows how tangled the problem has become.
To me, the new system looks like a mess. The goal is defensible because grades should tell students, parents, employers and graduate schools something real about performance. But an A cap, a hidden ranking system and fresh incentives for faculty are not a clean return to academic honesty. Harvard is trying to restore meaning to grades by adding another layer of machinery. That may be the best it can do, but it is still
machinery.
The larger point is that this problem did not begin with artificial intelligence. Harvard’s figures show that A grades rose from 24 percent in 2005 to 40 percent in 2015 and 60 percent in 2025. Purdue studied its own grading patterns in 2019 and found that average undergraduate course grades rose by 0.22 grade points between fall 2008 and spring 2017. Purdue was still below many peer institutions, but even Purdue could see the trend moving upward. Indiana University had been wrestling with grade inflation long before ChatGPT, with grade-distribution reporting dating back to faculty debates in the 1970s.
That is why blaming the latest technology or the latest generation of students misses the point. Grade inflation has been built into the modern university for decades.
Professors do not enjoy fighting students over grades. Departments compete for enrollments. Course evaluations matter. Parents are paying enormous bills and want results. Students are thinking about graduate school, internships and scholarships. In that environment, a B starts to feel like institutional punishment rather than an honest judgment.
The national pattern is hard to miss. By 2013, one widely used national grade-inflation database estimated that the average college GPA had reached about 3.15 and that 45 percent of letter grades were A’s. Purdue’s 2019 report listed peer GPAs that already looked generous: Duke at 3.51, Harvard at 3.65, Princeton at 3.39, Michigan at 3.37 and Indiana at 3.19. The numbers are not perfectly comparable, but the direction is obvious.
More Top Picks Best Laptop Cooling Stands For Students
The transcript was losing its sorting power before generative AI arrived.
AI now turns a weak grading system into a trap. A student can use AI to outline an essay, polish a lab report, debug code or manufacture a respectable answer in minutes.
Used properly, that can be a learning tool. Used lazily, it can make weak work look competent and competent work look polished. A 2026 Berkeley working paper found that courses with more AI-exposed tasks, including writing and coding, saw A grades rise by 13 percentage points after ChatGPT’s release, about a 30 percent increase relative to the 2022 baseline. That does not prove every student cheated. It does suggest the old grading bargain is breaking down.
This is where Harvard matters beyond Harvard. Elite colleges can absorb a lot of institutional nonsense because employers and graduate schools still recognize their names. Indiana, Purdue, Illinois, Michigan, Wisconsin, Texas, Florida and other public universities live closer to the practical world. Their students need grades to carry information into a labor market that is changing fast. Employers in an AI-driven economy care less about seat time and more about whether the student can write, calculate, code, reason and work with AI without becoming its clerk.
Universities now have to decide what a grade is supposed to certify. If grades certify effort, comfort and course completion, they will become less valuable. If they certify demonstrated competence under conditions that make AI use visible and measurable, they can still matter. That means more oral exams, in-class writing, supervised problem solving, project defenses, code walkthroughs and portfolio reviews. It also means universities must stop pretending that a transcript full of A’s automatically represents intellectual distinction.
Harvard’s reform may help, but it is not the final answer. It is a warning label. Once nearly everyone receives high marks, the institution has not made everyone excellent. It has made excellence harder to identify. AI accelerates that mistake because it makes surface competence cheap. The universities that remain relevant will be the ones that prove their students can still think when the machine is turned on.
Gerald Bradshaw is an international college admissions consultant with Bradshaw College Consulting in Crown Point.