AI grade inflation is the phenomenon, first measured at scale by Igor Chirikov (UC Berkeley CSHE, May 2026), in which students with access to AI receive systematically higher grades since ChatGPT's late-2022 launch, with the measured effect concentrated in AI-exposed academic disciplines. The proposed mechanism is that AI substitutes for the productive struggle through which learning occurs, so grades rise while underlying skill does not. It serves as the acquisition-side empirical anchor in the broader deskilling argument, paired with clinical-practice evidence on the maintenance side.
The Chirikov study
The empirical basis is a study by Igor Chirikov at the UC Berkeley Center for Studies in Higher Education, reported May 13, 2026, with the underlying paper hosted at eScholarship (Source: 'A' Grades Are Suddenly Everywhere Since the Arrival of ChatGPT (WSJ, May 13 2026) — Chirikov UC Berkeley study — WSJ Lindsay Ellis coverage).
The sample comprised more than 500,000 grades from 2018 to 2025 at a large Texas public university that publishes both syllabi and grade distributions. The method compared classes with more AI-exposed tasks (humanities, engineering, and writing- or coding-heavy courses) against classes less reliant on AI-substitutable work; before 2022, the study found little difference between the two groups.
AI-exposed classes gave approximately 30% more A's than non-AI-exposed classes after ChatGPT's launch, with corresponding declines in A-minus and B-plus grades. The effect was strongest in take-home and homework-heavy classes.
Chirikov attributes the pattern to AI displacing the effort required to learn: "AI may be helping people become more productive, to produce more, [but] I think it may harm their learning," and "Learning requires productive struggle that is eroded by AI."
Employer use of GPA as a hiring signal
Even as the reliability of GPA as a measure of skill has come into question, its use as a hiring filter has risen. The share of employers using GPA in hiring (per NACE) was on a downtrend in 2020, rose to 37% in 2023, and reached 42% in 2025. Handshake postings requiring a GPA of at least 3.5 rose from 9% in 2020 to 25% in 2025.
| Year | % of employers using GPA in hiring (NACE) | Handshake postings requiring ≥3.5 GPA |
|---|---|---|
| 2020 | (downtrend) | 9% |
| 2023 | 37% | (rising) |
| 2025 | 42% | 25% |
Chelsea Schein (Veris Insights) characterized employers as "talking from both sides of their mouth" — disliking AI use in applications while seeking AI fluency in hires. Barclays and Morgan Stanley maintain GPA minimums for some internship roles. (Source: 'A' Grades Are Suddenly Everywhere Since the Arrival of ChatGPT (WSJ, May 13 2026) — Chirikov UC Berkeley study)
Institutional response
Several universities reported on grading in response to the trend. A Harvard College report (February 2026) stated that "Employers have a legitimate interest in comparing students' performances, but current grading policies don't allow them to do so"; Harvard faculty voted in May 2026 on a proposal to cap the number of A's. A Yale University report (April 2026) stated that "Grades exist to communicate what students have learned. At Yale, as at many peer institutions, they no longer do." At Wharton, Schein shifted assessment weight away from homework, which can be substantially completed with AI, toward in-class quizzes and midterms. (Source: 'A' Grades Are Suddenly Everywhere Since the Arrival of ChatGPT (WSJ, May 13 2026) — Chirikov UC Berkeley study)
Relation to the deskilling argument
AI grade inflation provides the acquisition-side empirical evidence for the deskilling thesis, complementing maintenance-side evidence drawn from clinical practice. On the acquisition side, Chirikov's 500,000-grade dataset (May 2026) shows a roughly 30% increase in A-grades alongside a disconnect between grades and skill. On the maintenance side, Lancet Endoscopist Deskilling Study (2025) reported a 28.4% decline in adenoma detection rate after regular use of AI computer-aided detection (CADe). Taken together, the two findings are presented as evidence that AI use affects both the development of skill and its retention.
| Side | Empirical anchor | Effect |
|---|---|---|
| Acquisition (education) | Chirikov 500K-grade dataset, May 2026 | +30% A-grades, with grades disconnecting from skill |
| Maintenance (clinical practice) | Lancet Endoscopist Deskilling Study (2025) | -28.4% adenoma detection rate after regular AI CADe use |
Relationships
- supports: AI Deskilling, AI in Education, Education — AI Deployment
- instance-of: AI Labor Disruption (acquisition-side mechanism for credential erosion), LLM Fallacy (production-quality output mistaken for learned skill)
- depends-on: Jagged Frontier (uneven AI exposure across classes)
- contradicts: the framing that AI tools simply boost learning; the data indicate scoring inflation outpaces skill gain
- related: 'A' Grades Are Suddenly Everywhere Since the Arrival of ChatGPT (WSJ, May 13 2026) — Chirikov UC Berkeley study (anchor source), Igor Chirikov (created), 'I didn't want to be the guinea pig': inside tech's AI-fueled manager purge (Guardian, May 15 2026) (parallel "AI erodes signaling" thesis), Lancet Endoscopist Deskilling Study (2025) (clinical-side parallel), Chelsea Schein (Wharton, Veris Insights)
Sources
- 'A' Grades Are Suddenly Everywhere Since the Arrival of ChatGPT (WSJ, May 13 2026) — Chirikov UC Berkeley study — WSJ Lindsay Ellis coverage of the Chirikov UC Berkeley study (May 13 2026 print)
- Lancet Endoscopist Deskilling Study (2025) — Clinical-side deskilling counter-evidence (May 2026 ingest already covered)