Is AI to Blame for America’s Frozen Labor Market?

These days, something seems strange when you pass any university career fair. It’s not the tense atmosphere or the practiced pitches, but rather the dwindling number of recruiters. They used to be more numerous. When conversations do occur, they are usually more guarded and brief. Technically, businesses are still appearing. Simply put, fewer business cards are being distributed. Everyone from Silicon Valley CEOs to senators is blaming artificial intelligence for the shift in the American labor market over the last two years.

It should be noted that the panic is not wholly made up. There is a significant slowdown in progress. Some of the lowest job growth figures since the beginning of the pandemic were recorded in the final three months of 2025.

Goldman Sachs economists have been closely monitoring the trend, noting that the labor market “started to slow in the second half of 2025” after two separate shocks: a sharp slowdown in immigration from the record pace set in 2023 and 2024, and tariff uncertainty that made businesses cautious.

These are important structural forces. However, AI has also entered the discussion, and its role is more complex than the loudest voices will acknowledge, more difficult to quantify, and more difficult to discount.

According to former Transportation Secretary Pete Buttigieg, half of entry-level positions may disappear “not in 30 or 40 years, but in three or four.” Speaking at an Axios summit, Virginia Senator Mark Warner went so far as to forecast that unemployment among recent college graduates would increase from about 9 percent to 30 or 35 percent before 2028. The numbers are shared, amplified, and repeated. It’s an interesting tale. Additionally, it is mostly incorrect, or at the very least, it is based on dubious facts.

The Federal Reserve Bank of New York reports that in December 2025, the real unemployment rate for recent college graduates was 5.6%, not 9%. For the entire year, it had fluctuated between 5 and 6 percent—a little higher than the average for all workers, but far from crisis territory. Warner’s 35 percent forecast would necessitate a collapse in the graduate labor market that is about three times worse than what was observed in 2020, when the rate peaked at 13.4 percent during the height of COVID-19 shutdowns.

That does not imply that the worry is unfounded. Employment has obviously tightened for workers in their early twenties who had already made their way into tech and software roles. According to a 2025 Stanford University study, employment among people aged 22 to 25 had decreased by roughly 6% since ChatGPT’s launch in late 2022 in the professions most directly exposed to AI, with junior software development being the most obvious example.

Alongside that time frame, a revision from the Bureau of Labor Statistics revealed that the information sector had created 67,000 fewer jobs than previously thought during the previous 12 months. In a commentary, the chief economist at Comerica Bank stated that the updated data “more clearly” demonstrated AI replacing tech jobs. Even if the overall labor market hasn’t collapsed as the doomsayers predicted, it’s still possible that something genuine is occurring in a small portion of the white-collar industry.

The picture is made more nuanced by two different Federal Reserve surveys. According to research by the New York Fed, 12% of service companies that currently use AI said they had hired fewer employees in the previous six months as a result. Almost 25% of those who intended to use AI anticipated comparable decreases.

According to the Dallas Fed, roughly 10% of companies claimed that AI had reduced their overall need for employees. These aren’t trivial numbers. But they’re also not apocalyptic. And the same surveys noted something the headlines tended to skip: some firms were increasing hiring and others were actively retraining staff to work alongside the technology, not be replaced by it.

Dario Amodei, the CEO of Anthropic, offered perhaps the most striking vision of where this could lead — a world where cancer is cured, the economy grows at 10 percent annually, and 20 percent of people are still out of work. It’s a jarring combination, and it speaks to a genuine tension that even optimists haven’t fully resolved: productivity gains and job losses can coexist for years before the economy finds a new equilibrium. How difficult the change is, and for whom, is the question.

Early-career workers entering knowledge and content creation sectors are almost certainly going to feel it first, according to Goldman Sachs Research economist Joseph Briggs, who estimates that if AI adoption happens over a decade rather than in a compressed rush, the unemployment impact might add only 0.6 percentage points to the overall rate. That is doable. It becomes more difficult to absorb in a scenario that is more front-loaded.

There’s a tendency in this conversation to assume that “entry-level jobs” means “young people staring at laptops in open-plan offices.” That’s a particular and limited vision of American work. Most entry-level jobs are not knowledge jobs, because most jobs are not knowledge jobs. Fish and game wardens don’t write memos that AI can draft. Stonemasons, flight attendants, plumbers, home health aides, and the patrol officer walking a neighborhood beat at 2 a.m. are not about to be replaced by a chatbot, however persuasive the current generation of language models might be.

Even the more pessimistic scenarios for white-collar displacement would account for something like 2.6 million job losses per year — a real number, but one that sits alongside the roughly 20 million Americans who are laid off or fired under normal economic conditions every year anyway. The economy absorbs enormous labor market churn constantly. It might absorb this too, though with considerable pain concentrated among the people caught in the transition.

What may actually be happening is a more unglamorous story — a labor market slowed by tariff anxiety, reduced immigration, and cautious business sentiment, into which AI’s influence has crept quietly at the margins. Goldman Sachs projects unemployment edging up to around 4.5 percent in 2026, a modest increase driven by multiple factors simultaneously, not a single algorithmic villain.

Meanwhile, construction employment linked to data center buildout has risen by 216,000 jobs since 2022, as the infrastructure required to run the AI revolution generates its own demand for electricians, HVAC contractors, and lineworkers — trades that were never going to be automated away anytime soon. It’s an imperfect offset, and nobody pretending otherwise would be taken seriously. However, it adds complexity to the story of pure displacement.

There’s a feeling watching all of this unfold that the loudest predictions are doing something more than forecasting — they’re performing alarm in a way that shapes the political conversation without necessarily illuminating the economic reality. That’s not a reason to dismiss the concern. Young workers in tech and creative industries are navigating real uncertainty, and the data, however modest in scale so far, points in a direction worth watching carefully.

The honest answer to the question of whether AI is to blame for America’s frozen labor market is: partly, in specific sectors, against a backdrop of other forces doing at least as much work. The headline is unsatisfactory. However, compared to most of what has been spoken aloud, it is more accurate.