Irene Chang did everything she was supposed to do. She studied industrial and systems engineering at Georgia Tech, a degree she chose partly because it seemed safely employable, and started applying for entry-level data analyst roles nearly a year before graduating in May. Since last September she has submitted roughly 450 applications and landed about 19 interviews. So far, none has produced an offer.
"I feel like the entry-level skills that you have, they're important, but also something that AI can do very efficiently," Chang told NPR in a report published Tuesday. Her experience — and her suspicion that artificial intelligence is part of the problem — is increasingly common among the newest cohort of degree holders, and it is reshaping how young Americans think about work in the age of AI.
The Numbers Behind the Anxiety
The sense of a squeezed entry-level market is not imagined. According to the Federal Reserve Bank of New York, the unemployment rate for recent graduates — defined as 22-to-27-year-olds holding at least a bachelor's degree — stood at 5.7% as of June, well above the 4.1% rate for all workers.
The belief that AI is the culprit is widespread. In a ZipRecruiter survey cited by NPR, 47% of recent graduates said AI had already affected hiring in their field. Jacqueline Kline, 25, who finished a master's in communications at Florida State University this spring, told NPR she has applied to more than 500 entry-level jobs since December. "It's hard knowing that I'm doing everything that I can to be successful and feeling like that's not enough," she said.
The Case That AI Is to Blame
Stanford economist Erik Brynjolfsson gives the grievance real weight. "AI is not the whole story, but it's part of the story and the evidence is building," he told NPR.
Using payroll data, Brynjolfsson and his co-authors found that since late 2022 — when large language models like ChatGPT arrived — early-career workers aged 22 to 25 in AI-exposed roles such as software developers and marketing managers have experienced a 16% relative employment decline. Over the same period, employment for older workers in those same fields, and for workers of all ages in jobs that are hard to automate — home health aides, physical therapists, construction workers — held steady or grew.
Brynjolfsson identifies two mechanisms. Young workers always bear the brunt of hiring pullbacks, and junior roles are typically cut first. More distinctively, large language models are trained on codified knowledge: the written-down, textbook-style information that entry-level graduates rely on. The tacit, experience-based knowledge of senior employees is far harder for a model to replicate, so the newest workers are the ones competing most directly with the technology.
The Case That Something Else Is Going On
Harvard economist David Deming is not convinced. "If you look very carefully at the timing, it looks like the decline in junior hiring actually started a bit like six months before ChatGPT was released," he told NPR. "And so what that tells me is it's something else. I think it's more like remote work."
A recent New York Fed analysis supports that read: companies have become less likely to hire recent graduates into roles that can be done remotely, and as remote positions proliferated after the pandemic, unemployment among younger graduates rose in step. The analysis found that AI did not explain the increase — remote work did. Deming's logic is about training costs: junior employees require investment to bring up to speed, and doing that from afar is harder. Remote work also hands employers a national talent pool, making it easier to hire proven senior people instead of developing juniors.
University of Chicago economist Anders Humlum offers a third data point that cuts against the AI-displacement narrative. He points to a study by the financial accounting firm Ramp and workforce researchers Revelio Labs, which examined AI spending and headcount across more than 21,000 U.S. firms from early 2021 to early 2026. At companies making the largest AI investments, entry-level headcount grew 12% over the two years following AI adoption.
"If we look at the heavy users of these tools, the firms that are paying a lot of money to Anthropic and OpenAI to subscribe to their models, they are hiring more than anyone else," Humlum said.
A Shifting Consensus
The three economists disagree on how much damage AI is doing to early careers today, but all agree a major employment transition is underway. All are among the economists, executives and researchers who signed an open letter last month warning that AI could drive an economic transformation larger than the Industrial Revolution, with the potential for widespread job displacement.
Yet none of them counsels despair. Humlum says most students he talks to "just have the perception that they are being screwed over by the new technology" — a conclusion he calls very premature — and argues AI is positioned to help workers more than it hurts them. Brynjolfsson strikes a similar note: "This is a really tough time for a lot of the conventional jobs. I also think it's one of the most amazing times to be alive, where these technologies allow people to do things they never could have before."
For the hundreds of applicants like Chang and Kline still waiting for an offer, that reassurance may ring hollow. But the research suggests the story they are telling themselves about AI is, at minimum, incomplete — a mix of real technological displacement, a remote-work restructuring of junior hiring, and an ordinary cyclical squeeze landing all at once.
