Yes, by the figures that exist. In the second quarter of 2026 the New York Fed put unemployment among recent US college graduates at about 5.6 per cent and underemployment at 42 per cent. Payroll records show employment among 22 to 25 year olds in AI-exposed occupations well behind their less-exposed peers, while experienced workers show no comparable gap. Whether AI is the reason is contested: one large study finds the early-career hiring fall tracks remote work and not AI exposure, and the first direct test of graduate unemployment found no significant rise in summer 2026 relative to older graduates. The difficulty is measured, and the cause is still being argued.
The answer, in one line
By the available figures, yes. The New York Fed reported recent-graduate unemployment of about 5.6 per cent and underemployment of 42 per cent for 2026 Q2, and payroll data shows employment of 22 to 25 year olds in AI-exposed occupations 19 per cent below their less-exposed peers.
Definition#
AI-exposed junior employment gap: the difference between employment of 22 to 25 year olds in AI-exposed occupations and the level it would have reached had it kept pace with similarly aged workers in less-exposed occupations. Brynjolfsson, Chandar and Chen report it at 19 per cent in their August 2026 revision. The term is descriptive and is not a SuperSkills coinage.
Graduate unemployment is elevated and underemployment is the bigger number#
The Federal Reserve Bank of New York publishes quarterly labour-market figures for recent graduates, aged 22 to 27. The latest, for 2026 Q2, reads: "The unemployment rate stayed elevated at about 5.6 percent, and the underemployment rate edged up to 42 percent." The page describes conditions as challenging and offers no comparison with all workers in the text read here, so the headline says the market is tight for this group and nothing about why. The Bank does not attribute the figures to AI.
Payroll records show a gap that opens through hiring#
Brynjolfsson, Chandar and Chen of the Stanford Digital Economy Lab use high-frequency ADP payroll data covering millions of US workers through June 2026. Their August 2026 revision reports no evidence of widespread, economy-wide job displacement, and that employment of 22 to 25 year olds in AI-exposed occupations "now stands 19% below where it would be had it kept pace with that of their less-exposed peers; experienced workers show no comparable gap." The divergence operates "primarily through reduced hiring of young workers rather than increased separations", and the declines concentrate in occupations where AI usage mainly substitutes for human tasks. Where usage complements workers, employment is flat or rising.
The 19 is a gap between two groups of the same age. A fall of 19 per cent would be a different claim, and so would a gap against a historical trend. The design is observational, and youth hiring responds to interest rates and cohort size as well.
Hosseini Maasoum and Lichtinger at Harvard reach a similar pattern from a different source: résumé data covering 65 million workers at more than 280,000 firms. After a firm begins hiring people to integrate generative AI, its junior employment declines relative to non-adopters while senior employment trends stay largely unchanged, and the decline is "driven primarily by slower hiring rather than increased separations". It is an unrefereed working paper, and the authors describe the association as suggestive.
A rival explanation says the hiring fall follows remote work#
Lambert and Schindler, at Warwick and the LSE and at the Ellison Institute in Oxford, combined 243 million employer-employee linked records of new hires from Revelio Labs with 407 million online vacancy postings from Lightcast, across the US, UK, Canada and Australia from 2017 to 2025. By 2025, they report, a two-standard-deviation increase in an occupation's remote-work exposure predicted "a fall of around 5pp in the junior-share of new hires and around 3pp in the share of job ads requiring limited experience". Entering remote-work and AI exposure together, "the WFH effect remains, while the GenAI coefficient attenuates heavily and often becomes statistically insignificant."
The authors say they "do not interpret this evidence as ruling out strong impacts of GenAI on labor markets", and that it bears only on junior against senior hiring up to 2025. Remote work and AI exposure are strongly correlated across occupations, so separating them is hard, and each of the three studies above makes a different choice about how. Two other candidates, interest rates and post-pandemic over-hiring, are in the wider debate and are not tested in the studies read for this page.
Graduate unemployment did not jump in summer 2026, on the first direct test#
Fairlie and Wu of UCLA asked the narrower question directly: did unemployment among recent graduates rise when AI could first plausibly have reached them? Using Current Population Survey data from January 2022 to August 2026, for 22 to 25 year olds whose highest degree is a bachelor's, they report average June to August unemployment of 7.3 per cent in 2026, against 7.1 in 2022, 6.3 in 2023, 7.8 in 2024 and 7.2 in 2025. Compared with graduates aged 30 to 49, and with non-graduates aged 22 to 25, the summer 2026 change is small and not statistically significant, and it stays that way on a wider measure that adds graduates who want a job but have stopped searching.
Two limits matter for reading it beside the studies above. It measures unemployment, so a graduate who never got an offer and took other work does not show up, while the payroll and résumé studies measure hiring into particular occupations. And the authors say "neither the timing of adoption nor the available comparison groups provides a clean basis for causal interpretation", and that the results "do not rule out larger effects in the future". It is an NBER working paper, not yet peer-reviewed. Its 22 to 25 summer rate is a different measure from the New York Fed's 22 to 27 quarterly rate, so the 7.3 and the 5.6 do not conflict.
What the studies leave open for a person looking for a first job#
- The cause. The studies disagree on whether AI exposure or remote work explains the fall in junior hiring, and none isolates interest rates.
- Other countries and fields. The payroll and résumé data are American, and the New York Fed series covers US graduates only. Lambert and Schindler's four-country data is the one cross-national result here.
- The future. Every figure above describes 2022 to mid-2026. None forecasts the next cohort.
- Your occupation. The 19 per cent is an average across exposed occupations. Occupations where AI complements staff show flat or rising employment in the same data.
If the rung that trained people is the one disappearing, the first job costs more than the pay#
This section is interpretation, kept apart from the evidence above.
On 21 September 2025 Rahim Hirji wrote in The Missing Rungs: "we're not facing mass unemployment. We're facing what I'm calling Missing Rungs Problem." The studies above fit that shape better than they fit a story of general job loss: the same payroll data that finds no economy-wide displacement finds the gap in hiring at the bottom of exposed occupations. The argument that follows is about training. A first job taught people how to do work they would later supervise, and an employer that stops hiring for it saves the cost now and meets the bill in a thinner senior tier later. That second half is an argument and no study here measures it. The fuller treatment is on the missing rungs, and the question of how people become senior without the junior work is on how do juniors become senior.
Reading a job market for the rung, not the headline#
- Check whether AI substitutes or complements in your target role. The Stanford result separates occupations where AI usage replaces tasks from those where it supports workers. Reading job advertisements for what the junior is expected to do, and whether it is the task a model now does, is a practical version of that test. The method is an inference from the paper, not a tested job-search technique.
- Look for roles that still hire for development. Employers that run apprenticeships, rotations or supervised first-year programmes are keeping the rung. Ask what a first-year hire did in the last cohort.
- Build visible evidence of judgement. If routine tasks are the ones being absorbed, the part of a junior's work that still differentiates is checking, deciding and explaining. That is interpretation and no study here tests it. See should juniors use AI.
- Treat remote-first and in-person employers as different markets. If Lambert and Schindler are right, supervision cost matters, so the in-person end may behave differently. This is one working paper's finding and it is untested for any single employer.
Key sources
- Federal Reserve Bank of New York. The Labor Market for Recent College Graduates, 2026 Q2. Graded entry.
- Brynjolfsson, E., Chandar, B. and Chen, R. (2026). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford Digital Economy Lab, revised 12 August 2026. Graded entry.
- Hosseini Maasoum, S. M. and Lichtinger, G. (2026). Generative AI as Seniority-Biased Technological Change. SSRN working paper. Graded entry.
- Fairlie, R. W. and Wu, J. (2026). The Early Impacts of AI on Employment among Recent College Graduates. NBER Working Paper 35796, September 2026. Graded entry.
- Lambert, P. J. and Schindler, Y. (2026). The Broken Ladder: AI, Remote Work, and Early-Career Hiring. Discussion paper 228/26. Graded entry.
Related SuperSkills research#
On whether AI replaces the job outright, will AI replace entry-level jobs. On the employer side, should we hire fewer juniors. On the pattern itself, synthetic seniority. On the whole labour-market question, will AI replace my job.
About this research#
Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. The New York Fed page, the Stanford publication page, the Harvard abstract and the Lambert and Schindler paper were read at source on 3 October 2026; Fairlie and Wu was read in the NBER full text on 16 November 2026, when the New York Fed and Stanford pages were re-checked. All five are graded in the evidence base. AI-exposed junior employment gap is a descriptive term and not a SuperSkills coinage.
Evidence review · SS-2026-386 · Graded against the published rubric · How I use AI here · 2 working papers
Hirji, R. (2026). Is it harder to get a first job now?. The SuperSkills evidence base, SS-2026-386. https://thesuperskills.com/research/is-it-harder-to-get-a-first-job-now. Last reviewed 16 November 2026.
An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.
How citations and IDs work