Skip to content

All five questions

Will AI destroy more jobs than it creates?

The claim, as coded: “AI will displace more jobs than it creates”

The short answer AI-written

Will AI displace more jobs than it creates? Documented positions cluster in the middle: displacement and creation are both real, and the balance is contested. Early evidence shows concentrated losses in exposed work but no detectable aggregate employment effect yet, and new survey data show actual use differs from exposure scores.

A synthesis of the cited research, not a prediction or a representative consensus.

Why they disagree AI-written

Estimates of the share of jobs at high risk from automation range from about one in ten to nearly half, depending on whether whole jobs or single tasks are counted.

Each row is one assumption and the published papers that make it (of 154 on this question). Dots: their estimates, hover for the paper. Bar: where their findings point.

How much of each job can AI do?

A large share4 papers
1 for
A smaller share4 papers · middle estimate 11.5
Not read yet

Would settle it: Data on how the task content of individual jobs changes after firms adopt AI.

Will displaced workers find new work?

Yes, mostly11 papers · middle estimate 7.25
3 mixed, 5 against
Not easily5 papers · middle estimate 32.5
1 mixed, 1 against

Would settle it: Tracking workers in exposed occupations over several years: where they work and what they earn.

Does AI replace workers or work alongside them?

Replaces11 papers · middle estimate 32.5
1 for, 4 mixed
Works alongside, or both32 papers
No estimates
17 mixed, 4 against

Would settle it: Firm-level data on hiring, headcount and task mix before and after AI adoption.

What kind of evidence is the estimate built on?

Exposure scores12 papers · middle estimate 11.5
5 mixed
Observed changes37 papers
No estimates
1 for, 9 mixed, 5 against

Would settle it: Comparing exposure forecasts made in earlier years with the employment changes that followed.

The case made for the claim

Early market evidence drives the agree side: Del Rio-Chanona documents generative AI already reducing demand for exposed freelance work, and Bana finds deterioration in AI-exposed jobs beginning even before ChatGPT. Prettner's search-and-matching models find displacement pressures that policy does not automatically offset, and Korinek's distributional analysis takes significant displacement and wage pressure seriously. Bengio has repeatedly flagged large-scale labor disruption among advanced AI's key risks. Korinek and Suh's task-based scenarios show wages can fall if automation outpaces capital accumulation, and collapse if full automation is reached with bounded task complexity.

Still open

  • Will new-task creation keep pace with displacement this time, given AI targets cognitive rather than routine manual work?
  • Are the early freelance-market losses a leading indicator for employment generally, or a niche effect?
  • How long is the lag between task-level capability and measurable aggregate employment effects?
  • How closely does measured workplace use of generative AI track the exposure scores most displacement estimates rely on?

Sources

  1. New Frontiers: The Origins and Content of New Work, 1940–2018
  2. Winners and losers of generative AI: Early Evidence of Shifts in Freelancer Demand
  3. Automation and New Tasks: How Technology Displaces and Reinstates Labor
  4. The Turing Trap: The Promise & Peril of Human-Like Artificial Intelligence
  5. The Turing Transformation: Artificial Intelligence, Intelligence Augmentation, and Skill Premiums
  6. Artificial Intelligence and Jobs: Evidence from Online Vacancies
  7. Scenarios for the Transition to AGI
  8. What Work Does Generative AI Do?
  9. GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models

The numbers

Forecasts of jobs at risk from AI

Each point is one estimate of the share of jobs at high risk of automation or expected to be lost, by the date it was published; bars show the author’s own range. 8 estimates from 7 sources, grouped like the growth forecasts: economists and AI insiders. Definitions differ (high risk of automation, displaced by a date, a slice of jobs such as entry-level white-collar work), and none of them is a net count: most also expect new jobs. Read the spread, not a single number.

Economists: 8–1501020304050602013201420152016201720182019202020212022202320242025% of jobsFrey & Osborne 2013 (economist): 47%. US jobs at high risk of computerisation. “According to our estimate, 47 percent of total US employment is in the high-risk category”Frey & Osborne 2013OECD 2016 (Arntz et al.) (economist): 9%. Jobs at high risk of automation, average of 21 OECD countries. “Overall, we find that, on average across the 21 OECD countries, 9 % of jobs are automatable.”OECD 2016 (Arntz et al.)McKinsey 2017 (economist): 15%. Global workforce that could be displaced by automation, 2016-2030. “around 15 percent of the global workforce, or about 400 million workers, could be displaced by automation in the period 2016–2030”McKinsey 2017OECD 2018 (Nedelkoska & Quintini) (economist): 14%. Jobs highly automatable (over 70% probability), OECD countries in PIAAC. “About 14% of jobs in OECD countries participating in PIAAC are highly automatable”OECD 2018 (Nedelkoska & Quintini)WEF 2025 (economist): 8%. Current jobs displaced by 2030 (all trends, not only AI; net growth 7%). “the displacement of the equivalent of 8% (or 92 million) of current jobs”WEF 2025Goldman Sachs 2025 (economist): 6.5% (range 6 to 7). US workforce displaced if AI is widely adopted. “could displace 6-7% of the US workforce if AI is widely adopted”Goldman Sachs 2025Amodei 2025 (AI insider): 50%. Entry-level white-collar jobs AI could wipe out. “AI could wipe out half of all entry-level white-collar jobs” Implied: "half" read as 50%Amodei 2025Amodei 2025 (AI insider): 15% (range 10 to 20). Unemployment rate AI could cause. “spike unemployment to 10-20% in the next one to five years”Amodei 2025

EconomistsAI insidersShaded: the middle half of each group

All 18 numbers from 15 papers, in 3 charts, each with its quote

Who says what

Documented positions only, coded with cited evidence by an AI agent: 26 of 400 mapped researchers so far. Everyone else is unknown: no reviewed evidence, which is neither agreement nor neutrality.

6 agree11 it's complicated9 disagree26 of 400 coded

Agree (6)

It's complicated (11)

Disagree (9)

Economists and computer scientists

Documented positions on this claim of researchers on the co-authorship map, by field (from their OpenAlex topics). Researchers without a documented position are not counted.

FieldAgreeIt’s complicatedDisagree
Economists (241 on the map)6 of 249 of 249 of 24
Computer scientists (51 on the map)Too few to say (2 coded)

Each position with the researcher’s reasoning and cited evidence

What to read

Essentials on jobs

Read

Know

Follow

  • David Autor leans disagreeEconomist, Massachusetts Institute of Technology

    Author of 'Why Are There Still So Many Jobs?', on how automation has both removed and created work.

  • Erik Brynjolfsson leans disagreeEconomist, Stanford University

    Co-author of the payroll study on early-career workers in AI-exposed jobs.

  • Kinder Futures (Molly Kinder)Newsletter · 2 posts in this guide

    A newsletter on AI and the workforce from a researcher who studies the future of work.

All essentials on jobs, for 10 minutes to an afternoon

Where the papers point

5 agree46 it's complicated14 disagree65 of 313 coded

How each published paper’s own finding bears on the claim, read from its title and abstract by an AI model (AI-written). 65 of the 313 published papers address this question; the rest are not counted. This codes findings, not authors.

Findings that support the claim (5)

and 2 more in the research library

Findings that it depends (46)

and 43 more in the research library

Findings that go against the claim (14)

and 11 more in the research library

All research on this questionNews on this question