The First Rung Is Narrowing: What the Data Say About AI and Graduate Jobs

In September 2026, a National Bureau of Economic Research working paper put one of the year's loudest labor-market fears to what its authors call a useful first test. Robert Fairlie and Jane Wu of UCLA used Current Population Survey microdata for June, July and August 2026 and found that unemployment among recent college graduates did not spike, whether measured against earlier summers, older graduates or young workers without a degree^[1][16].

Among bachelor's degree holders aged 22 to 25 who were no longer enrolled in school, the summer unemployment rate averaged 7.3 percent. That is higher than in three of the four previous summers, which ranged from 6.3 percent in 2023 to 7.8 percent in 2024^[16]. Counting graduates who want a job but are not searching adds nearly two percentage points, and that broader measure showed no significant rise either^[1].

The forecast had been grim. In March, BlackRock chief executive Larry Fink said he worried that this year's graduates could face their highest unemployment rate in years, even without a recession^[10]. So the first rung of the career ladder has not snapped. But payroll evidence suggests it is narrowing in specific places, which is the harder problem for universities.

What the payroll records show

In August, the Stanford Digital Economy Lab revised its study Canaries in the Coal Mine, by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen. Using ADP payroll records covering millions of US workers through June 2026, the authors report that 22-to-25-year-olds in AI-exposed occupations now hold 19 percent fewer jobs than they would if their employment had tracked that of less-exposed peers. Experienced workers show no comparable gap^[3].

Between November 2022 and June 2026, employment of that age group fell about 11 percent in the two most exposed fifths of occupations and grew about 10 percent in the three least exposed. On the same measure, the shortfall was 15 percent on July 2025 data. It comes mainly through reduced hiring, not more departures, and is concentrated where AI is used to automate tasks instead of assisting workers^[3].

“The entry-level effects we’re measuring are real, persistent and widening”

Brynjolfsson made the remark in a Washington Post interview, as quoted by Ars Technica, and added that he had grown more worried about an economy that holds employment steady while shutting newcomers out^[4].

The authors are candid about limits. The paper finds no economy-wide job displacement and calls its facts descriptive, not causal. The pattern weakens when occupations' education levels are controlled for, shows some divergence that predates generative AI, and is stronger in ADP's sample than in national surveys^[3].

Why two careful studies disagree

They measure different things. Payroll data count jobs in particular occupations. The unemployment rate counts graduates who are looking for work and have none^[2]. By our own reasoning, a graduate who misses a junior analyst post and takes other work would lower the first measure and leave the second untouched. The Federal Reserve Bank of New York's tracker is consistent with that reading: in the second quarter of 2026, unemployment among recent graduates was about 5.6 percent, while 42 percent were underemployed, in jobs that typically do not require a degree^[5].

AI is also not the only suspect. Research published by the New York Fed in June estimates, by its authors' rough calculation, that remote work can explain 64 percent of the rise in unemployment among young graduates between 2017-19 and 2022-24, because training newcomers at a distance is harder, and notes that the rise began before generative AI spread^[6]. Harvard economist David Deming made the same timing argument to NPR, saying junior hiring began to fall roughly six months before ChatGPT's release^[7]. The Stanford paper reports that its gap persists after controlling for remote work and interest rates^[3].

Firm-level data cut the other way. NPR reported that a Ramp and Revelio Labs study of more than 21,000 US firms found that at the companies investing most in AI, entry-level headcount grew 12 percent in the two years after adoption^[7]. TechCrunch noted that the sample leans toward fast-growing, tech-forward firms, and quoted its authors as saying it does not show that AI universally creates jobs^[8]. An October 2025 analysis by the Yale Budget Lab and Brookings likewise found no discernible disruption in the broader labor market in the 33 months after ChatGPT's launch^[9].

Job-posting data look worse, and need care. UK jobs site Adzuna listed 8,383 graduate vacancies in July 2026, down from 15,397 a year earlier, yet Charlie Ball of Jisc cautioned that firms may simply be labeling fewer roles as graduate roles^[11]. On Handshake, the campus recruiting network that raised $200 million at a $3.5 billion valuation in 2022, postings fell more than 16 percent in the year to August 2025 while applications per role rose 26 percent, Fortune reported^[10][14].

The exposed asset is codified knowledge

For educators, the most useful finding is why juniors are exposed. The Stanford authors propose that generative AI substitutes for codified knowledge, the formal, documented kind taught through courses and textbooks, and complements tacit knowledge built through practice and mentorship. Their data show slower entry-level employment growth in occupations heavy in codified knowledge, though they stress that the estimates are not causal and overlap with education levels^[3].

That is uncomfortable for universities, whose core product is codified knowledge. The New York Fed's evidence points the same way: inside one Fortune 500 company, feedback and mentoring fell away sharply when colleagues were physically separated, and younger workers lost the most^[6]. Whether the cause is AI or distance, the scarce good is the same: supervised practice.

What universities should change

• Curricula: Grace Lordan, associate professor in behavioral science at the London School of Economics, argues that many degree programs are still training students for the routine analysis AI can now do. She calls for placements built into degrees, projects designed with employers and AI fluency as a core requirement, not an elective^[12].
• Work-integrated learning: Strada and the Burning Glass Institute found in 2024 that 52 percent of US graduates are underemployed a year after graduation, and 73 percent of those who start out underemployed remain so a decade later. Graduates with at least one internship had 48.5 percent lower odds of underemployment^[13]. Investors have backed the apprenticeship version: Multiverse, which embeds salaried apprentices with employers alongside its own coaches, raised $220 million at a $1.7 billion valuation in 2022^[15].
• Assessment: Lordan cites a 2026 Institute of Student Employers survey in which 43 percent of respondents said AI had changed entry-level roles without any formal redesign, and more than a third said new graduates fell short of expectations in adapting to the workplace^[12]. If the first job now begins where the routine draft ends, programs should assess what follows it: checking AI output, judging incomplete evidence and explaining a decision.

September's result is no all-clear. Fairlie and Wu warn that the classes of 2027 and later may be more affected if workplace AI use keeps deepening^[2][16]. Deming expects a rough transition: "it's going to be very bumpy and messy"^[7].

The WAC view

The alarm overstates what the data show, and the reassurance understates it. Graduates are not being shut out of work, but hiring into junior roles in AI-exposed fields is weakening, and the knowledge most exposed is the kind universities teach most. Institutions that wait for economists to settle causation will be late. Supervised practice, through placements, apprenticeships and assessed real projects, is the sound response whether the culprit proves to be AI, remote work or the business cycle.

What to do with this

The events behind this article, in WAC News

• Handshake raises $200 million Series F at $3.5 billion valuationJanuary 19, 2022
• Multiverse raises $220 million Series D at $1.7 billion valuationJune 8, 2022

Sources

1. The Early Impacts of AI on Employment among Recent College Graduates (Working Paper 35796), Robert W. Fairlie and Jane Wu. National Bureau of Economic Research, September 2026.
2. AI was supposed to hit new grads hard. So far, unemployment data says otherwise.. Ars Technica, September 25, 2026.
3. Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence (revised August 2026), Erik Brynjolfsson, Bharat Chandar and Ruyu Chen. Stanford Digital Economy Lab, August 12, 2026.
4. AI is hitting entry-level jobs hardest, Stanford study finds. Ars Technica, August 24, 2026.
5. The Labor Market for Recent College Graduates (2026:Q2 highlights). Federal Reserve Bank of New York, 2026.
6. Remote Work Leaves Younger Workers Sidelined, Natalia Emanuel, Emma Harrington and Amanda Pallais. Federal Reserve Bank of New York, Liberty Street Economics, June 1, 2026.
7. Many recent grads say AI is making it harder to get a job. Economists aren't so sure. NPR, August 18, 2026.
8. The AI jobs debate just got messier. TechCrunch, June 29, 2026.
9. Evaluating the Impact of AI on the Labor Market: Current State of Affairs. The Budget Lab at Yale (with Brookings), October 1, 2025.
10. BlackRock CEO Larry Fink warns AI is creating a ‘crisis’ for Gen Z workers: the class of 2026 could face the highest unemployment in years—even without a recession. Fortune (via Yahoo Finance), March 18, 2026.
11. Graduate job vacancies drop by almost 50% in a year, survey suggests. BBC News, August 24, 2026.
12. AI means fewer graduate jobs – how to make sure they still exist in the future (Grace Lordan). LSE British Politics and Policy, September 23, 2026.
13. Talent Disrupted (report page and executive summary). Strada Education Foundation and The Burning Glass Institute, February 21, 2024.
14. Handshake Expands to Help Young People Build Relationships with Employers and Launch Careers. Chan Zuckerberg Initiative, January 19, 2022.
15. Multiverse nabs $220M at a $1.7B valuation to expand its tech apprenticeship platform. TechCrunch, June 8, 2022.
16. The Early Impacts of AI on Employment Among Recent College Graduates, Robert Fairlie and Jane Wu. CESifo (Working Paper No. 12994; full text of the paper also issued as NBER Working Paper 35796), September 2026.

Written by the WAC editorial team. Facts are drawn from the numbered sources above and were checked against them before publication; opinions appear only under “The WAC view”. The organisations and people named are not affiliated with the World Assessment Council unless stated. Send a correction