The internet is abuzz with speculation. AI is the prevailing topic at the moment. However, the decline in unemployment — in particular, among college grads — has implied a loose association with the rise of AI. Have entry level jobs dropped because of ChatGPT? Is AI really making a significant impact on the jobs economy since its launch 3 years ago?
One thing is clear: U.S. Unemployment among youth (ages 16-24) is the highest it’s been in four years. Since February 2021, where youth unemployment stood at 10.9%, August 2025 rates, according to the Federal Reserve Bank of St. Louis, were sitting at 10.5% unemployment. In comparison, the overall U.S. unemployment rate was 4.3% in August 2025.
In March 2023 Goldman Sachs predicted: AI will destroy or degrade 300 million full-time jobs. This came on the heels of OpenAI’s launch of ChatGPT and DALL-E, which they touted, at the time, as a capability that is able to “generate content that is indistinguishable from human-created output and to break communication barriers between humans and machines…” Goldman Sachs’ research, leveraging data on job tasks in the U.S. and Europe, estimated two-thirds of existing jobs would be exposed to AI automation, for which generative AI could conceivably substitute up to one-fourth of them.
Fast forward to present day, Harvard published a paper, Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data, which concluded, “Following adoption, junior employment declines sharply in adopting firms relative to non-adopters, while senior employment remains largely unchanged. The junior decline is concentrated in occupations most exposed to GenAI and is driven by slower hiring.”
Stanford Study released in the same period, Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Their study, which sought to determine effects of labour markets exposed to GenAI concluded: given the “widespread adoption of generative AI, early-career workers (ages 22-25) in the most AI-exposed occupations have experienced a 13% relative decline in employment.”
Jing Hu is a seasoned researcher, data analyst and the author of Second Order Thinkers. Her recent analysis, Are Junior-Level Jobs Really Killed by AI? questioned the results of the Harvard paper. She added, “Can a 62-million data study make a wrong and dangerous assumption?” This Harvard research covered the period from 2015-2025 and included resume data from 62 million workers from 285,000 firms as well as 200 million job postings on Linkedin.
According to Hu, the study revealed that the “junior hiring crash started in 2022 before many people knew what prompt engineering was. In March 2022, an inflation surge, caused by Pandemic supply and demand imbalances, was a time when the Federal Reserve started hiking interest rates 25 basis points from near-zero to 0.25-0.5%. But that was just the beginning.” She adds, “Over the next 16 months, the Fed executed 11 consecutive rate hikes, driving rates from 0% to 5.25-5.50% by July 2023, ending the longest period of cheap money in modern history.” Hu emphasizes the cumulative rate increase of 5.25% in over a year was the “most aggressive tightening campaign in decades.”
The timeline for AI as a causal instrument is misleading, as she notes, “The junior hiring collapse started in Q1 2023, right as the rate shock hit its stride, months before most companies had even figured out what to do with Generative AI and LLMs. AI is likely a convenient excuse later, adding marginal pressure only from late 2024 onward, long after the real damage was done.”
Harvard’s study also analyzed LinkedIn job postings and flagged terms like “ChatGPT,” “prompt engineering,” and “retrieval-augmented generation”—identifying GenAI integrator roles that represented just 0.066% of all postings. Hu highlights one example: a Junior Product Manager role that, despite being labeled “junior,” required “experience integrating with GenAI security and safety products”—not something a recent graduate would have. Hu points to the study’s definition of “‘GenAI integrator’, while most had only heard about GenAI the same year, was already filtering out true entry-level positions.”
In Economic Downturns, A Consistent Pattern that Emerges
Hu is adamant, “Every economic shock, every time… the same pattern. Junior hiring craters. It stays depressed. And when it comes back, if it comes back, it never returns to the old baseline.”
She alludes to the historical economic shocks 1990-91 and 2001 Dot-bomb. In both cases, high unemployment resulted and persisted while the economy recovered. After the Dot-bomb of 2001, Hu adds, “what followed was a prolonged and severe contraction in employment. Entry level positions disappeared as companies retained senior staff, eliminating junior roles.” According to the Economic Policy Institute, this 2001 bubble burst was marred by a slower recovery, leading to a “tougher economy for highly educated workers, and resulting in large declines in employment opportunities for new graduates.” In the downturns between 1979-2002 the report concluded that entry level jobs had been disproportionately hurt by the weak labor market.”
Earlier this year, The Economist’s analysis on this very topic in their piece, “Why AI hasn’t taken your job, questioned the imminent job displacement ushered in by GenerativeAI. They point out that “young graduates’ relative unemployment started to rise in 2009.”
Between 2022 – 2025, Hu reports that tech job posts “fell 36% below pre-pandemic levels as of mid-2025.” However, postings for roles with less than one year of experience experienced a more significant hit of 50% between 2019 and 2024.
A recent Economist podcast confirmed that post-Pandemic hiring was significantly above pre-Pandemic levels. Tech hiring in 2021 and 2022 was two times the hiring levels. A similar trend followed white collar jobs in professional services and information services, for which college grads are typically hired for — the very professions that are seeing more stagnant growth today. Indeed, the job market platform, “noted that the earlier hiring boom, broader economic conditions, and interest in AI could explain this year’s crash in demand for tech workers.”
The Macro Events Omitted from the AI-Kills-Jobs Studies
These macro policies have wielded a blow to employment opportunities, and were variables that did not appear in the Harvard study.
Hu also disproved two other claims from the Harvard study. In Q1 2023, a few months after ChatGPT first launched, the study saw junior employment drop 9% while senior employment climbed. Second, when comparing roles between firms adopting AI versus those that had not, there was a 10% relative drop in junior hiring, with the study claiming, “Beginning in 2023Q1, the coefficients drop sharply, reaching roughly a 10 percent decrease after six quarters.”
The timeline of events exposes a significant gap. As per Hu,
- March 2022: The fed starts hiking rates from 0%
- November 2022: ChatGPT launches
- Q1 2023: Harvard sees the “AI impact” begin
What the study failed to acknowledge is the impact of the rate hike, the confounder, which had an enormous effect on the employment rate.
Hu creates the scenario of two organizations:
1) Company A was more vulnerable to rate hikes (venture-funded, high debt, dependent on cheap money).
2) Company B was less vulnerable to rate hikes (profitable, cash reserves, no debt dependency).
She argues that the study assumes that “firms posting AI jobs in 2023 are different because they have adopted AI,” when, in fact “they can very likely be two different beasts.” So while both types of organizations are exposed to the Fed’s rate hikes, Company A, which is highly debt dependent, will more likely cut junior staff. The added complexity is nuanced in each company’s relative risk to the current cost of borrowing. Hu concludes the study failed to consider “enough variables to claim these junior jobs issues have nothing to do with the economy.”
The Federal Reserve had already started hiking interest rates by March 2022. By July 2023, interest rates had soared from near zero levels to 5.25-5.50%. These rate hikes were meant to slow demand, and slow investment. The hiring boom at 0% interest rates made sense. At the 5.5% rate, companies will reduce costs, and that translates into potential layoffs.
Hu mentions that since 1980 every major economic shock has resulted in “ 1) disproportionate cuts to entry level hiring 2) permanent downward shifts in youth labor force participation 3) long-term wage scarring for affected cohorts 4) failure to recover to pre-shock hiring levels even during expansions.”
She doesn’t discount the eventual impact AI will have on these jobs, but “inventing a new story” each time a downturn occurs whether it be “off-shoring in 2001 and automation in 2008, “ or AI in 2025… the real culprit consistently can be attributed to “economic cycles and managerial cowardice.”
Hu’s concern from the Harvard, Stanford study, World Economic Forum and MIT — who have all published their versions of “AI killed junior jobs” research — is that readers give credence to their brand and take them at their word. She adds, “To be fair, this [Harvard] team did lay out the caveats in the body of their paper. They admit their definition of “adoption” is a blunt tool, their timing is fuzzy, and they can’t fully rule out other confounders, like big tech firms suddenly finding religion about costs after bing-hiring in 2021-22. But mass media, pundits and LinkedIn influencers do not read footnotes.” By their very name, she adds, “the institutional brands make the claim stick, whether or not the correlation holds up.”
State of AI in 2025: AI Adoption is Poor
For AI to have an indelible impact on the job market, let alone be a disruptor on junior employment, organizations need to firmly adopt. The recent McKinsey Report on the State of AI in 2025 claims that the technology is still very much in the early stages, and “Nearly two-thirds of respondents say their organizations have not yet begun scaling AI across the enterprise.. In addition, the curiosity with agents remains just that as 62% indicate they are still experimenting with the technology.
The report indicates that most organizations are still at the stage of value creation, with only 39% reporting EBIT impact, “ most have not yet embedded them deeply enough into their workflows and processes to realize material enterprise-level benefits.” In the coming year, the outlook for AI varies with 32% expecting decreases, while 13% increases in the use of AI. The remainder see no change.
MIT’s similar report on the State of AI in 2025 references the ‘GenAI Divide’, highlighting the “billions of dollars invested in enterprise GenAI pilots” that have failed to yield value. There have been more ebbs in the journey of LLMs than significant milestones achieved, and with the billions in circular deals that the hyperscalers, cloud providers and AI chip providers are spending on each other, the illusion of massive revenue is enough to overvalue each entity. Pundits believe this AI bubble is going to burst.
In the meantime, speculation of recession looms. The Trump tariffs have set in motion a domino effect that have increased operational costs, resulting in sudden shifts in consumer spending, and have compounded the unemployment drama. Data from the October Jobs Report from Challenger, Gray and Christmas shows U.S. employers announced 153,074 job cuts, up 183%% from September. The reasons for the October U.S. job cuts: “Some industries are correcting after the hiring boom of the pandemic, but this comes as AI adoption, softening consumer spending, and rising costs drive belt-tightening and hiring freezes.”
Time will tell whether AI will shatter the future of work and wreak havoc in a global economy that is highly dependent on labour. That inevitability will be a future state. In the meantime, breathe a sigh of relief that Harvard got it wrong this time.