The entry-level job market in mid-2026 is best captured in one word: frozen, not empty. Employers still have roles open, but they fill them slowly, and the people already in jobs are staying put, so the whole market has stopped churning. It is harder than the 2021-22 hiring boom, softer than the pre-pandemic years, and nowhere near the wreckage of 2009-10. The feeds say no one is hiring. The government’s openings data complicates that: vacancies are still high even as hiring stays subdued, which looks less like an empty market than a cautious one that has stopped moving. The opposite claim, that the market is basically fine and your search is the only broken part, misses how long people now stay stuck once they fall out of work.
Is anyone actually hiring entry-level workers?
Probably yes, though the openings data answers this only for the economy as a whole. In the May 2026 JOLTS report, the Bureau of Labor Statistics counted 7.6 million open positions, an openings rate of 4.6%, high by recent standards. Those are job vacancies counted on the last business day of the month, not online postings. And JOLTS carries no seniority or occupation breakdown, so it cannot say how many of those openings are first jobs. What it does show is the shape of the market.
The same report shows hiring running slow. Hires, counted as a flow over the whole month, ran at 5.2 million, a rate of 3.3%. Openings are counted at a moment in time and hires across the month, so the gap between them is not a conversion rate, and reading it as one claims more than JOLTS can show. Still, high vacancies sitting alongside subdued hiring is at least consistent with a cautious, low-churn market: the roles are on the books, and employers are in no hurry to fill them. Quits point the same way. At 3.1 million, a rate of 1.9%, they sat low by recent standards, which means people who already have jobs are staying put. Fewer people leaving generally means fewer openings created to backfill them, and less overall churn in the system.
Employer hiring plans tell a similarly measured story, though they are plans, not hires. In its spring update, the National Association of Colleges and Employers reported that employers expected to hire about 5.6% more new graduates from the Class of 2026 than the year before. That is a projection from an employer intention survey, not observed hiring, so treat it as a signal of direction. It points up, and it follows two straight years of weak markets for new graduates. It does not line up cleanly against the boom-era projections of 2021-22, because NACE now reports this figure as a median rather than the mean it used before, specifically to stop a few outliers from skewing it, so the old double-digit numbers and today’s 5.6% are not measuring the same thing.
The June jobs report fills in the human side. Overall unemployment was 4.2%, and payrolls grew by 57,000 on the month, a soft but positive number. As for how long people have been looking: the median unemployment duration was 11.0 weeks, down from 11.6. That figure is the time the currently unemployed have been searching so far, across all ages and all workers, not a finished stopwatch on how long a job hunt takes and not specific to new graduates. The freeze shows up more clearly among those who fall out of work for a long stretch. People unemployed 27 weeks or longer made up 27.3% of the jobless, about 1.9 million, up 286,000 from a year earlier.
Is grad underemployment as bad as it sounds?
This is the number most likely to be quoted at you without its context. The New York Fed’s most recent reading, for the first quarter of 2026, put underemployment among recent college graduates at 41.5%. It gets read as an unemployment or hardship number. It measures something much narrower.
The Fed counts a recent graduate as underemployed when they are working in a job that typically does not require a bachelor’s degree. What it looks at is the kind of job someone holds, not whether they have one or what it pays. A graduate waiting tables is underemployed by this definition, and so is a graduate in a salaried role that has not historically demanded a degree. So the figure says nothing directly about whether a graduate is unemployed or how much they earn.
The context that rarely travels with the number is its history. Recent-graduate underemployment has sat in the low 40s for decades, on the New York Fed’s own series, and the current reading is down slightly from where it sat at the end of 2025. This is a chronic feature of how young workers enter the labor market, one that long predates the current anxiety about it.
There is a real cost nearby, though it is not this statistic. Graduating into a weak labor market leaves a mark on early earnings, whatever job you land. The well-known study by Oreopoulos, von Wachter, and Heisz, which followed cohorts who left school into recessions, found an initial pay penalty of around 9% that faded to roughly zero over about a decade. That scar is about the timing of your graduation, not about taking a job below your degree level, and it lands on early earnings, which start lower and then recover over roughly the following ten years.
Have recent grads lost their edge?
Mostly yes, and it is the genuinely new fact in this whole picture. Recent-graduate unemployment has recently moved above the unemployment rate for the overall workforce, a reversal of the usual pattern that economists at the St. Louis and Cleveland Feds have flagged. In a matched comparison from the St. Louis Fed, over the first seven months of 2025, recent graduates aged 23 to 27 averaged 4.59% unemployment, above the overall rate of 4.18%. In 2019 the same comparison ran the other way, graduates at 3.25% against 3.67% overall; that ordering has flipped.
Two qualifications keep this off the doom pile. In that same matched set, graduates still did better than their same-age peers without a degree, who averaged 6.38%. And the levels are low by any longer historical yardstick. A separate New York Fed series, covering a slightly different age band, 22 to 27, puts recent-graduate unemployment at about 5.7% in the first quarter of 2026. For scale, overall unemployment today, at 4.2%, is nowhere near its roughly 10% peak in 2009-10. The degree still helps; what it lost is the near-automatic advantage over the average worker that graduates once took for granted.
Is AI to blame for the entry-level slowdown?
The viral version of this story is that AI ate the entry-level job. The primary research does not support that as a statement of fact, and the researchers closest to the data openly disagree with each other.
Set four prominent analyses side by side:
- Stanford’s “Canaries in the Coal Mine” (November 2025), built on ADP payroll data, found that workers aged 22 to 25 in the most AI-exposed occupations saw a 16% relative decline in employment. That figure is relative, measuring a shrinking share of jobs next to less-exposed peers. It does not mean one in six young workers lost a job.
- A Harvard SSRN working paper by Hosseini Maasoum and Lichtinger, “Generative AI as Seniority-Biased Technological Change,” found that at firms it infers to have adopted AI, junior employment fell by about 7.7% within six quarters, and the decline came from slower hiring rather than layoffs. That is a within-firm effect at inferred adopters, not a count of jobs vanishing across the economy. Much larger percentages circulate attached to this paper; they describe reductions in the hiring rate on bases that get reported inconsistently, so we cite the paper’s clearest number and leave the viral one alone.
- The New York Fed, analyzing Lightcast job-posting data (May 2026), found limited evidence of a distinct AI-driven decline in junior hiring. Junior and senior postings have largely moved in parallel, and the divergence people blame on AI started before 2022.
- A second New York Fed paper (June 2026) estimated that remote work explains about 64% of the post-pandemic rise in young-graduate unemployment. That is a back-of-the-envelope estimate comparing 2017 to 2019 against 2022 to 2024, a window that ends before the 2025-26 rise this piece is mostly about, so it speaks to the earlier post-pandemic climb rather than the latest slowdown.
Put together, the defensible read is narrow. AI is associated with softer hiring in the occupations and firms most exposed to it. It has not been established as the primary cause of the broad entry-level slowdown, and several serious Fed economists attribute more of the earlier rise to remote work and the general freeze. The market clearly got harder for young graduates; whether AI is the reason remains unproven.
What this means if you are job hunting right now
The frozen-market read changes what to expect from a search. A cautious, low-churn market is slow by nature, not because your application is uniquely broken, so a first search that stretches on for weeks is closer to the current normal than a sign you are doing it wrong.
Given that, the higher-leverage move is usually to go narrow and warm rather than to widen the net. Pick a handful of roles you genuinely fit, and ask someone inside each company for an introduction. A warm introduction will not conjure a job that is not really there, but it is the kind of nudge that counts for more when hiring is slow and cautious. It also helps to set a fixed follow-up cadence, for example a check-in a week after applying and again a week later, so a slow no does not quietly turn into a silent one you never chased down.
Keeping that structure straight as the applications stack up is the part of a slow search that quietly falls apart, and it is what we built CareerPigeon to handle, down to a Chrome extension that saves applications straight from your Gmail. And if you want a single number to watch for the market loosening, watch quits: when people start leaving jobs again, the backfilled openings and the churn a first-time job seeker depends on come back with them.
How we know this: every figure here traces to a primary source. The openings, hires, quits, and duration data come from the Bureau of Labor Statistics, specifically the May 2026 JOLTS report and the June 2026 Employment Situation; JOLTS measures job vacancies rather than online postings, and carries no seniority split, so we have not used it to count entry-level roles. The hiring-plans figure is a forecast from the National Association of Colleges and Employers’ spring update, an employer intention survey rather than observed hiring. Recent-graduate underemployment comes from the Federal Reserve Bank of New York’s college labor market data, and the crossover in unemployment against the overall workforce is documented by the St. Louis and Cleveland Feds; we have kept the graduate-versus-overall-versus-nongraduate comparison inside a single matched St. Louis Fed series rather than splicing sources. On AI’s role we set the primary studies against one another rather than picking the scariest: Stanford’s ADP-based “Canaries in the Coal Mine” (November 2025), a Harvard SSRN working paper by Hosseini Maasoum and Lichtinger on seniority-biased technological change, and two New York Fed analyses (job postings, drawing on Lightcast data, May 2026; remote work, June 2026). The earnings-scar figure, on graduating into a weak market, is from Oreopoulos, von Wachter, and Heisz. Where a number is contested among serious researchers, or is a latest-quarter reading that will move, we have said so in the sentence that uses it.