Why you have seen much bigger numbers
Start with every role figure reported for a cut the press linked to AI. Keep only the ones where the company itself named AI and stated the number. Then weight what survives by how strong the evidence is. Each step is a filter you can click into.
1 · Reported in AI-linked cuts
Every figure reported for these cuts.
2 · Employer named AI
3 · Evidence-weighted
As an independent cross-check, Challenger, Gray & Christmas separately counts 121,401 US job cuts that employers attributed to AI, year to date. It is a different method (announced US cuts, not our evidence-weighted worldwide register), so the two figures measure different things and are not expected to match.
Year by year, since 2023
Since 2023, employers have named AI in the loss of 101,916 roles worldwide, 31,779 after evidence-weighting. In 2025 alone they named AI for 13,314, while Challenger counted 54,836 US job cuts attributed to AI. Each year below is counted by the same three steps. Open a year to see its events.
Why there is no single number
Three distinctions decide whether a job cut belongs in an "AI replaced it" count. Most trackers blur all three. Keeping them apart is what makes the figure defensible.
- Attribution: who said AI did it?
- There is a real difference between an employer naming AI in a filing and a journalist inferring it. Only cuts the company itself attributed to AI count toward the verified total; a cut that is merely alleged and cannot be verified is held out, not quietly discounted into the number.
- Execution: did it actually happen?
- An announcement is a plan; an executed cut is a person out of work. We record which is which and never present a planned reduction as a completed one.
- Counted versus merely reported.
- Of the roles reported this year, 175,499 are in cuts the press linked to AI that do not meet the evidence bar. They are shown alongside the count, not inside it, so the headline is never inflated by figures we cannot stand behind.
Is it getting worse?
Counting totals answers "how many so far." To answer "is the pace rising or cooling," the AI Layoffs Index reads 59.4 (Elevated) and is currently trending down. It is deliberately scaled against AI's own history since 2023, not against total unemployment, so it measures how today's AI-attributed cutting compares with AI's own past, not what fraction of all jobs are affected. Whether AI is showing up in unemployment itself is a separate question: see what the jobs data shows about AI and unemployment. The work AI creates is a separate count again: the hiring radar on AITraining.jobs, a data service we also built, tracks every role the AI-training platforms post, with what it pays and which experts it wants.
Which jobs, and which companies?
- What jobs will AI replace?: the jobs in the cuts employers themselves tied to AI, counted from the register by job family.
- Are companies blaming AI for layoffs?: every cut grouped by what the employer itself said, and the ten largest AI-linked cuts.
- Which companies have cut jobs for AI: every counted event, company by company, with its evidence and source.
- The full timeline of events: filterable by company, sector, evidence tier, and attribution.
Frequently asked
How many jobs has AI replaced so far?
The most defensible answer is a range, not one figure. Since 2023, employers have named AI in the loss of 101,916 roles worldwide, of which 31,779 hold up after each cut is weighted by how strong its evidence is; in 2026 so far the two figures are 72,018 and 20,844. Independently, Challenger, Gray & Christmas counts 193,226 US job cuts attributed to AI from May 2023 through October 2026, 121,401 of them this year. The register behind these numbers holds 345 AI-linked job cuts since 2023, each linked to its source. The confirmed, employer-attributed total is in the tens of thousands; the millions you sometimes see are projections of AI exposure, not counts of jobs already gone. Current through October 2026.
See the live countHow many jobs were lost to AI in 2025?
In 2025, employers themselves named AI for 13,314 roles worldwide: the figures they stated in the cuts where they gave AI as a reason, among the 79 AI-linked job cuts the register holds for that year. Of those roles, 4,433 hold up after each cut is weighted by how strong its evidence is. All 79 cuts together reported 137,056 roles, counting figures the employer did not tie to AI or did not state itself. Challenger, Gray & Christmas, counting announced US cuts by its own method, recorded 54,836 job cuts attributed to AI in 2025.