A study of firms in 41 countries is being shared online as proof that AI is shutting young people out of work. The paper behind it, from Stanford University and King’s College London, is more careful. It finds that companies adopting generative AI took on more senior staff, while the drop in junior staff was small and not statistically significant.
Spotted via unusual_whales (@unusual_whales) on X.
What the study is
Working paper, not peer reviewed. “How Does AI Change Labor Demand? Evidence from 41 Countries”, by Bharat Chandar (Stanford) and Bouke Klein Teeselink (King’s College London), is dated 20 September 2026 and posted on SSRN [1][2]. King’s publicised it on 6 October [3].
The authors analyse “1.25 billion job postings and 154 million employment records across 41 countries”, using Revelio Labs data from January 2021 to March 2026 [1]. (The King’s release says 158 million [3]; TSN uses the paper’s figure.) They count a firm as an AI adopter when its job ads involve using generative AI, and they study foreign branches of multinational companies, comparing them with similar branches in the same country and sector [1].
What it found
Authors’ estimates. The paper’s exact wording: “senior employment rises by 6.7 percent, while junior employment is 2.5 percent lower, an estimate that is not statistically significant” [1].
Because seniors grew, the junior share of staff at adopting branches was “1.9 percentage points lower than in control affiliates by March 2026” [1]. Total employment “goes up by a marginally significant 3.3 percent”, significant only at the 10% level [1]. For a typical branch of about 255 positions, the point estimates imply “about 8 more seniors” and “about 4 fewer juniors” [1].
“Junior” here means Revelio’s two lowest seniority levels, entry and junior, not an age group [1]. Across 31 countries with enough data, the junior share fell in 23, with statistically significant falls in seven, including the US, Brazil and Saudi Arabia [1].
Chandar summed it up in the King’s release: “generative AI is changing the mix of jobs within organisations rather than reducing overall employment” [3].
Correcting the viral figure
Posts on X, including the one where TSN spotted this, said junior employment fell “-3%” [5]. The paper says 2.5% lower, and not statistically significant [1]. The “3” may come from elsewhere in the paper: the 1.9-point share drop “is equivalent to 3.3 percent of that baseline share”, and juniors were 2.9% lower after one year, an estimate the authors call “somewhat imprecise” [1]. The King’s release itself says junior employment “falls by 2.5%” without the significance caveat [3].
For a UK view, see TSN’s look at the ONS business survey and a Banque de France study of young workers’ hiring (related, publishing soon).
What this does not prove
- That AI cut junior jobs. The junior estimate is not statistically significant [1].
- That AI creates jobs overall. The 3.3% total rise is only “marginally significant”, and the authors show firm-level gains and economy-wide changes “can have opposite signs” [1].
- Anything about young people as such. “Junior” is job seniority, not age [1].
- Effects at every employer. The sample is foreign branches of multinationals, and adoption is inferred from job ads [1].
The Bottom Line
The paper’s evidence points to firms reshaping their workforces towards senior staff after adopting AI, not to a measured collapse in junior jobs. Senior employment rose 6.7%; junior employment was 2.5% lower, not significantly. It is a working paper, not yet peer reviewed.
Related on TSN: AI and Jobs: What a UK Survey, a French Study and One Layoff Actually Show; AI Is Eliminating Wall Street Jobs — And Banks Are Finally Admitting It
Sources
- Bharat Chandar and Bouke Klein Teeselink, “How Does AI Change Labor Demand? Evidence from 41 Countries”, working paper dated 20 September 2026, Stanford Digital Economy Lab PDF (not peer reviewed; abstract, introduction and section 5 read directly). https://digitaleconomy.stanford.edu/app/uploads/2026/09/HowDoesAIChangeLaborDemand.pdf
- SSRN listing, abstract 7498743 (working paper; the page blocked TSN’s fetch on 9 October, so the paper was read from the Stanford copy). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7498743
- King’s College London, “AI adoption linked to increase in employment globally”, news release, 6 October 2026 (says 158 million records and “between 2021 and 2016”, which appears to be a typo; the paper covers January 2021 to March 2026). https://www.kcl.ac.uk/news/ai-adoption-linked-to-increase-in-employment-globally
- Bharat Chandar, “AI’s impacts on jobs around the world”, What’s Next (Substack), author’s commentary. https://bharatchandar.substack.com/p/ais-impacts-on-jobs-around-the-world
- unusual_whales (@unusual_whales), X post, 9 October 2026, 13:38 BST (read via the X API). It goes beyond the evidence: it gives junior employment as “-3%” (the paper says 2.5% lower, not statistically significant), adds “over 5 years” (the estimates run to March 2026, from a panel starting January 2021), frames it as “not for young workers” (the paper measures seniority, not age) and says AI “is creating jobs” (the overall rise is only marginally significant). Its 1.9-percentage-point share figure matches the paper. https://x.com/unusual_whales/status/2108537461739585995
 finds that at AI-adopting firms in 41 countries, senior employment rose 6.7% while junior employment was 2.5% lower, "an estimate that is not statistically significant". Viral posts said junior jobs fell 3%. Here is what the paper says, and its limits.){kind=link}