Generative AI could expose the equivalent of 300 million full-time jobs worldwide to automation
Assessment
Evidence favors the claim, but the chain is incomplete or the sources are secondary.
The figure originates in Goldman Sachs Research's March 2023 analysis of generative AI and economic growth, which mapped AI capabilities onto detailed occupational task data and extrapolated the resulting exposure shares to global employment. The estimate rests on two quantitative premises, that roughly two-thirds of U.S. occupations are exposed to some degree of AI automation and that a quarter to half of exposed occupations' workload could be automated, together with the assumption that task-based analysis of occupational databases can meaningfully estimate AI exposure.
Independent evidence corroborates the order of magnitude: the IMF's 2024 cross-country analysis found about 40 percent of global employment exposed to AI, and the ILO's generative-AI index finds broad partial exposure, with transformation of jobs more likely than wholesale displacement. Goldman Sachs itself reaffirmed the estimate in 2026 while noting that AI is also likely to create jobs, particularly in power and data-center infrastructure. No credible analysis denies that generative AI could expose work on this scale.
The number should nonetheless be read as an order-of-magnitude model output, not a count. Multi-model comparisons show that exposure classifications vary widely depending on which AI model scores the tasks, more recent work documents a persistent gap between what AI could theoretically do and what it is observed doing in practice, and the extrapolation from U.S. and European occupational data to world employment adds further uncertainty. The assessment would strengthen if independent full-time-equivalent-weighted global estimates converged on the same range, and weaken if the workload-share premise were found unsound or replications diverged from 300 million by an order of magnitude.
Full reasoning: the evidence and decisions behind this verdict
Periodic re-examination roughly a week after the prior verdict; the question was whether the evidence landscape has moved. It has not moved materially.
What was checked. (1) Whether Goldman has revised or retracted the estimate: it has not. The 2026 labor-market article (www.goldmansachs.com/insights/articles/how-will-ai-affect-the-us-labor-market) and Goldman's own March 2026 public restatement (x.com/GoldmanSachs/status/2034640957992267982) both reaffirm the 300 million figure, now recorded among the claim's affirming instances. (2) Whether new credible denials or order-of-magnitude divergences have appeared: none found. Recent criticism continues along the lines already priced into the verdict, targeting the meaning and precision of exposure scores rather than the claim's modest "could expose" assertion. Notable recent strands: analyses of Anthropic usage data showing a capability-deployment gap (tasks scored as fully exposed that AI is not observed performing in practice), and a February 2026 ILO research brief (www.ilo.org/sites/default/files/2026-03/Research%20Brief_Workers%20exposure%20to%20AI.pdf) noting that exposure measures rest on static task lists. Both reinforce the caveats attached to the framework assumption that task-based analysis can meaningfully estimate AI exposure without producing a credible negation of the claim.
Subclaim standing is unchanged since the prior pass: the two-thirds occupation-exposure share remains supported with convergent independent estimates; the quarter-to-half workload share remains unassessed and is the softer premise, since the 300 million count scales directly with it. All recorded instances affirm; no denying instances exist.
Verdict unchanged: supported at credence 0.75. Supported rather than verified because the figure is a counterfactual model output whose methodological framework is under live dispute and one load-bearing premise is unassessed; supported rather than contested because the dispute qualifies precision, not the claim's truth as stated. What would change it: assessment of the workload-share premise as contradicted, or independent FTE-weighted global replications diverging from 300 million by an order of magnitude.
Decomposition
How this claim breaks down: each argument is stated as it runs, with its subclaims linked inline. ↗︎ opens a subclaim; the map shows how they fit together.
Because roughly two-thirds of U.S. occupations are exposed to some degree of AI automation and a quarter to half of exposed occupations' workload could be automated, applying these shares to employment worldwide yields on the order of 300 million full-time-equivalent jobs exposed to automation. The extrapolation takes U.S. and European occupational exposure patterns as a usable basis for global employment, and it assumes that task-based analysis of occupational databases can meaningfully estimate AI exposure.
Granting its premises, the arithmetic goes through: applying the exposure and workload shares to global employment yields a figure on the order of 300 million full-time-equivalent jobs. The argument currently leans hardest on the quarter-to-half workload share, which is not yet assessed and with which the count scales directly, while the two-thirds occupation-exposure share is supported by convergent independent estimates. Two caveats qualify the inference itself: the extrapolation from U.S. and European data to world employment adds uncertainty the premises do not capture, and the framework assumption that task-based analysis can meaningfully estimate AI exposure remains under live methodological dispute, most recently over the gap between theoretical task exposure and observed AI use, so the conclusion is best read as an order-of-magnitude estimate rather than a count.
Provenance
Where this claim has been said, linked to its canonical form.
Shifts in workflows triggered by these advances could expose the equivalent of 300 million full-time jobs to automation
A new wave of AI systems may also have a major impact on employment markets around the world.
Goldman Sachs Research estimates that 300 million jobs globally are exposed to automation by AI.
A 2026 follow-up on AI and the U.S. labor market that reaffirms the 2023 estimate while noting AI is also likely to create jobs, particularly in power and data-center infrastructure buildout.
According to Goldman Sachs Research, 300 million jobs globally could be exposed to AI automation over the next decade. However, AI is also likely to help create jobs—particularly in the buildout of the power and data center infrastructure required to sustain the boom
Goldman Sachs' official account restating its research estimate while promoting its 2026 analysis of AI and the U.S. labor market, adding that AI is also likely to create jobs in power and data-center infrastructure.
Assessment history
0 status changes over 2 assessments. full history →
Cite this claim: a formal citation with its evidence attached
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
The attention this claim received was paid for by a funded mandate. Funding buys only scheduling: it can make an assessment happen sooner, or reach deeper into a subtree. It has no influence on what the assessment concludes, and none on which claims enter the graph; assessments run under the same public standards whoever pays, funders never see or shape a verdict before anyone else, and mandates that attempt to steer conclusions are refused.
Created by extractor · Aug 10, 2026. Every judgment on this page is accompanied by a reasoning trace.