Key Takeaways

- A grades jumped 13 percentage points in writing and coding courses after ChatGPT launched
- The grade increase appears only in homework, not proctored exams, suggesting outsourced work
- Oral presentation grades didn't change, a placebo test that supports the AI-replacement theory
AI grade inflation is real, and the pattern points to students outsourcing work rather than learning better. A UC Berkeley study analyzing more than 500,000 grades at a large Texas research university found that courses heavy on writing and coding assignments saw A grades jump 13 percentage points after ChatGPT launched in November 2022. The spike shows up in homework. Exam scores haven't moved.
That gap is the study's central finding. If AI tools genuinely helped students learn, you'd expect better performance across all assessments. Instead, the grade boost clusters in unsupervised assignments where students can hand off work to ChatGPT without detection. Proctored exams, where that option disappears, show no comparable increase.
What the 500,000-grade dataset actually shows
Researcher Igor Chirikov tracked grade trends across eight fall semesters, from 2018 through 2025, in 319 courses spanning 84 departments. Each course's AI exposure was measured by its assignment mix from fall 2022 syllabi, before ChatGPT existed. The logic: courses that assigned more writing and coding, where AI performs best, should show stronger effects if AI is doing the work.
They did. The share of A grades rose by about 30 percent above the 2022 baseline in high-exposure courses. Average GPA climbed 0.12 points. The grade distribution narrowed as A-minus and B-plus grades got bumped up to straight A's.

The homework weight test is what makes the case. In courses where homework counted for more than the median share of the final grade, A's rose by an extra 16 percentage points compared to courses below the median with the same AI exposure. In those lower-homework courses, the effect was small and statistically insignificant.
Chirikov ran a placebo test too. Oral presentation assignments, where AI is far less useful, showed no grade changes. That result is difficult to reconcile with broad learning gains or sorting effects alone, he writes.
Why this matters for employers and graduate programs
Grade inflation isn't new. At Harvard, the share of A grades climbed from 24 percent in 2005 to 60.2 percent in 2025. Earlier research blamed teaching evaluations that reward leniency, competition between universities, and institutional grading policies. But those drivers kicked in at the grading stage, after students had turned in their work.
AI works differently. It changes how the work gets made, before instructors ever see it. If grades in writing and coding courses increasingly reflect AI-generated output rather than real skills, employers and graduate programs face worse selection decisions. The GPA loses its signal value.
Chirikov flags a feedback loop: if AI takes over skill-building tasks during college, graduates could end up weaker in exactly the areas where AI is strongest. That could accelerate automation and widen skill gaps in the job market.
Sam Altman admits the education system hasn't adapted
OpenAI CEO Sam Altman recently said that three and a half years after ChatGPT's launch, the education system has barely responded to AI. He expected a year of cheating followed by systemic overhaul. Instead, he can't point to any meaningful systemic change. Without one, he warns, critical thinking skills risk significant atrophy.
Altman still believes education can adapt, the way it has through earlier technological leaps. But he argues some skills, like writing and coding, should keep being taught because they train the mind itself. I'm someone who thinks through writing, and I write a lot of things I never show anyone, but it's still important to me for figuring things out, Altman says. People say the same thing about programming.
Norway recently banned AI tools from most elementary schools and limited their use in secondary schools. Prime Minister Jonas Gahr Stoere said uncritical AI use tempts students to skip important learning.
What can instructors actually do?
Moving everything to proctored exams isn't enough, and it isn't simple. The study suggests designing assignments that either limit AI use or fold it in deliberately. Documentation of the work process, follow-up interactions that prove understanding, or oral defenses of written work could help.
Some institutions are experimenting with AI-resistant assignments: work that requires local knowledge, personal reflection, or iterative revision visible to the instructor. Others are integrating AI explicitly, asking students to critique or improve AI-generated drafts.
Neither approach scales easily. Both require rethinking course design from the ground up.
Logicity's Take
For AI product teams, this study is a reminder that usage metrics don't equal value creation. ChatGPT adoption in education is massive, but if the measurable outcome is inflated grades rather than improved learning, that's a product problem. Teams building AI tools for education or enterprise training should consider how to measure genuine skill transfer, not just task completion. The homework/exam divergence Chirikov found is exactly the kind of signal that distinguishes tools people use from tools that actually help.
Frequently Asked Questions
How much did grades increase after ChatGPT launched?
In writing and coding heavy courses, the share of A grades jumped 13 percentage points, about 30 percent above the 2022 baseline. Average GPA rose by 0.12 points.
Why do researchers believe this is outsourced work rather than better learning?
The grade increase appears only in homework, not proctored exams. If AI improved learning, gains should show up everywhere. The effect also disappears in oral presentation assignments where AI can't help.
What is the proposed solution to AI grade inflation?
The study suggests redesigning assignments to either limit AI use or integrate it deliberately, with documentation of work processes or follow-up interactions that prove understanding.
Has any country banned AI in schools?
Norway recently banned AI tools from most elementary schools and limited their use in secondary schools, citing concerns about students skipping important learning.
Altman's views on AI development and critical thinking in education
Need Help Implementing This?
Building AI tools for education or enterprise learning? Logicity covers the research, product patterns, and regulatory shifts shaping this space. Subscribe for weekly analysis on AI product strategy.
Source: The Decoder / Matthias Bastian
Huma Shazia
Senior AI & Tech Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
Related Articles
Browse all
Bezos AI Lab Gets $10B: What Project Prometheus Means
Jeff Bezos is closing a $10 billion funding round for Project Prometheus, an AI lab focused on physics-based AI for manufacturing and engineering. With a $38 billion valuation and backing from JPMorgan and BlackRock, this signals a major shift in enterprise AI investment toward industrial applications.

Kimi K2.6 Open-Weight AI: 300 Agents at a Fraction of the Cost
Moonshot AI's Kimi K2.6 matches GPT-5.4 and Claude Opus 4.6 on coding benchmarks while running 300 parallel agents. For businesses locked into expensive API contracts, this open-weight model could slash AI infrastructure costs while delivering enterprise-grade automation.




