A new study tracking nearly 320 courses across eight semesters at a major US university has produced stark evidence that artificial intelligence is driving a 13-point surge in grade inflation, primarily by replacing student work on unsupervised homework assignments rather than enhancing genuine learning. The research, which examined 319 courses across 84 departments from fall 2018 through fall 2025, found that the share of A grades jumped by 13 percentage points\u2014roughly 30 percent above the pre-AI baseline\u2014while the average GPA climbed by 0.12 points and the overall grade distribution narrowed significantly, with A-minus and B-plus grades increasingly being bumped up to straight A’s.<\/p>\n
The study\u2019s design is what gives its findings particular weight. By measuring each course\u2019s AI exposure through its assignment mix, derived from fall 2022 syllabi written before ChatGPT existed, the researcher isolated the precise variable: the proportion of writing and coding tasks, the very areas where large language models perform best. The analysis then tracked grade changes semester by semester, creating a clean before-and-after picture of AI\u2019s impact on academic outcomes.<\/p>\n
The central research question was whether these higher grades reflected actual learning gains or simply AI doing the work for students. To find out, the study\u2019s author, Igor Chirikov<\/a>, examined how much weight homework carried in each course\u2019s final grade. The logic is straightforward: if AI genuinely improved learning, grade increases should appear consistently whether a course leaned heavily on homework or on proctored exams. But if AI was merely replacing student effort on unsupervised assignments, the effects should cluster in courses where homework carried more weight.<\/p>\n The data confirms the second scenario decisively. In courses where homework counted for more than the median share of the final grade, the rise in A grades was an additional 16 percentage points compared to courses below that median with the same AI exposure. In the lower-homework courses, the effect was small and not statistically significant. As Chirikov writes, this result is \u201cdifficult to reconcile with broad learning gains or sorting effects alone.\u201d A placebo test strengthens the conclusion: for oral presentation assignments, where AI is far less useful, grades did not budge.<\/p>\n Grade inflation at US universities is not new. The study notes that at Harvard, the share of A grades climbed from 24 percent in 2005 to 60.2 percent in 2025, driven by factors like teaching evaluations that reward leniency, competition between universities, and institutional grading policies. But Chirikov argues AI operates on a fundamentally different level. Every earlier driver of grade inflation kicked in at the grading stage, after students had already turned in their work. AI changes how the work itself gets made before instructors ever see it\u2014it rewrites the production process, not just the evaluation outcome.<\/p>\n The implications extend far beyond campus. If grades in writing- and coding-heavy courses increasingly reflect AI-generated output rather than genuine student skills, employers and graduate programs could make systematically worse selection decisions. The study also flags a dangerous feedback loop: if AI takes over the skill-building tasks that happen during college, graduates could emerge weaker in precisely the areas where AI is strongest. That dynamic could accelerate automation and widen skill gaps in the job market, creating a workforce that relies on AI but cannot critically evaluate its output.<\/p>\n The study does not leave the problem open-ended. Moving everything to proctored exams is not sufficient, nor is it practical at scale. A more effective approach, Chirikov suggests, is designing assignments that either deliberately limit AI use or intentionally fold it into the learning process\u2014for instance, through documentation requirements that force students to show<\/a> their work, or through follow-up interactions that prove genuine understanding. The key is making sure that the grade reflects what the student can actually do, not just what an AI assistant can produce on demand.<\/p>\n Even OpenAI CEO Sam Altman<\/a>, in a recent interview, acknowledged that the education system has barely responded to AI three and a half years after ChatGPT\u2019s launch. He had expected a year of cheating followed by systemic overhaul. Instead, he can point to no meaningful structural change\u2014and without one, he warns, critical thinking skills risk what he called \u201csignificant atrophy.\u201d Altman still believes education can adapt, as it has through earlier technological leaps, but he emphasizes that some skills, like writing and coding, should remain core curriculum because they train the mind itself. \u201cI\u2019m someone who thinks through writing,\u201d Altman said, \u201cand I write a lot of things I never show anyone, but it\u2019s still important to me for figuring things out.\u201d<\/p>\n The practical takeaway from this study is immediate and actionable. For educators, the fix is not to ban AI tools outright but to redesign assessment so that the final grade actually measures human understanding, not machine output. That means incorporating process documentation, verbal follow-ups, or proctored components that test the skills students are supposed to be learning. For hiring managers and graduate admissions, it means recognizing that an A in a writing-heavy course may no longer carry the signal it once did, and that direct assessment\u2014through work samples, interviews, or coding challenges\u2014is now more important than ever. For developers and technologists, this study serves as a clear warning: as AI<\/a> tools become more embedded in daily workflows, the critical thinking and foundational skills that come from doing the work without AI assistance must be intentionally preserved, or the next generation of graduates may find themselves dependent on tools they never fully learned to use.<\/p>\n","protected":false},"excerpt":{"rendered":" A new study tracking nearly 320 courses across eight semesters at a major US university has produced stark evidence that artificial intelligence is driving a 13-point surge in grade inflation, primarily by replacing student work on unsupervised homework assignments rather than enhancing genuine learning. The research, which examined 319 courses across 84 departments from fall […]<\/p>\n","protected":false},"author":7,"featured_media":84527,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/cards.overcentral.com\/cards\/en\/57536.png","fifu_image_alt":"AI Drives 13-Point Grade Inflation as Students Outsource Work","footnotes":""},"categories":[349],"tags":[],"class_list":["post-57536","post","type-post","status-publish","format-standard","has-post-thumbnail","category-articles"],"fifu_image_url":"https:\/\/cards.overcentral.com\/cards\/en\/57536.png","fifu_image_alt":"AI Drives 13-Point Grade Inflation as Students Outsource Work","_links":{"self":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/posts\/57536","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/comments?post=57536"}],"version-history":[{"count":0,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/posts\/57536\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/media\/84527"}],"wp:attachment":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/media?parent=57536"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/categories?post=57536"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/tags?post=57536"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}What This Means for Employers and Graduate Programs<\/h2>\n
A Concrete Solution: Rethinking How We Design and Evaluate Assignments<\/h2>\n
What This Means for Developers, Educators, and Hiring Managers<\/h2>\n