Analysis of more than 500,000 grades revealed that “A” marks increased by 30% in courses subjected to AI, raising concerns about “task displacement” and the diminishing value of college degrees.
Analysis of more than 500,000 grades revealed that “A” marks increased by 30% in courses subjected to AI, raising concerns about “task displacement” and the diminishing value of college degrees.
In his working paper, “Artificial Intelligence and Grade Inflation,” Chirikov examined more than 500,000 student enrollments in 319 courses across 84 departments at a significant Texas research institution between 2018 and 2025.
The results are startling: after ChatGPT became available, the biggest grade gains were seen in courses that largely relied on AI-vulnerable tasks like composing essays and coding.
Students who were previously receiving C marks are suddenly receiving A grades, Chirikov told Axios.
Quick Facts
| Key Detail | Information |
| Lead Researcher | Dr. Igor Chirikov (UC Berkeley) |
| Data Size | 500,000+ student enrollments |
| Courses Analyzed | 319 courses in 84 departments |
| Key Finding | 30% increase in A grades in AI-exposed courses |
| Most Affected Fields | Writing-intensive courses, coding courses |
How Researchers Isolated AI’s Impact on Grades
Chirikov had a problem: he was unable to arbitrarily grant ChatGPT access to certain courses while excluding others.
In order to compare grade trends in more AI-exposed courses vs less AI-exposed courses before and after ChatGPT’s release, he employed a difference-in-difference research design.
“AI-exposed” courses: what are they?
with a higher proportion of assignments that AI can successfully replace student labor, mainly essay writing and coding jobs completed as unsupervised homework.
“Less AI-exposed” courses: what are they?
Sculpting, lab work, live oral presentations, and class debates are examples of courses that AI finds difficult to imitate.
The Placebo Test
Chirikov used the percentage of oral presentations in a course as the exposure measure in a placebo test to validate his findings.
If his primary conclusions were about AI task substitution, this metric shouldn’t predict grade gains
because live oral presentation performance cannot be replaced by current AI technologies.
Chirikov clarified, “The interpretation that the grade increases are specific to course tasks where AI
capabilities are strongest is supported by the placebo test, which found no significant effect on grades.”
Why This Matters for Employers: The Signaling Problem
Employers use Grade Point Averages (GPAs) as crucial signals. GPA is now used by 42% of businesses to select applicants, up from 37% in 2023.
The issue is that the signal becomes less trustworthy if grades increase because AI is enhancing submitted work rather than because students are learning more.
Chirikov warns: “My study shows that AI may make this problem worse by inflating grades most in the courses where AI can most easily be substituted for student effort.”
Employer Responses
Some companies are already responding:
Company Action
Barclays ‘ minimum GPA requirement for internships
Morgan Stanley GPA thresholds for certain roles
General trend Nearly 25% of job postings requiring GPA now ask for 3.5 or higher (up from 9% in 2020)
However, even prestigious universities are raising concerns. According to a Yale University report:
“The purpose of grades is to communicate what students have learned. At Yale—and at many peer institutions—they no longer serve that function.
In a similar vein, Harvard’s February 2026 research said that businesses find it challenging to compare pupils due to the present grading methods.
Chelsea Schein, an adjunct professor at Wharton and vice president of Veris Insights, pointed out the dilemma that employers encounter:
“They make contradicting claims. Employers are interested in graduates who are proficient with AI technologies, but they are wary of applicants who use the technology while applying.
Leave a Reply