This page reads its numbers from
data/ when it opens. If you opened it from disk (file://),
serve it over HTTP — python -m http.server — or build the standalone copy with
python scripts/build_standalone.py.
The AI layoff story, checked against the record
How many jobs is AI really taking?
Every month, layoffs are blamed on AI. We went looking for the companies that actually said so — on the record, with a headcount. So far in …:
The story surged. The record never followed.
Announced U.S. job cuts citing AI, month by month — and the part of each that meets the bar. Under each month: AI's share of all announced cuts.
The seam, accumulated
Running totals for the year. The space between the lines is the distance between the story and the record.
View as a table
One layoff. Several numbers.
A figure changes as it travels — from filing to memo to tracker to headline. When our sources disagree we keep every figure and say which one we use. Nothing is averaged, because an average of two claims is a third claim nobody made. The most instructive cases first; a state WARN count is only ever a partial.
Who actually said it.
Every major 2026 layoff we've checked — … so far — grouped by how much of the “AI” is actually on the record. Click a row for the reasoning. Every source is linked; open them yourself.
Zoom out. What the government actually counts.
The Bureau of Labor Statistics measures how many people are laid off or discharged every month, across the whole economy. It never says why. It is the yardstick the AI story sits inside.
Layoffs & discharges, every month since 2022
Seasonally adjusted, all U.S. nonfarm employers · BLS JOLTS.
How we count.
- Blamed on AI
- Challenger, Gray & Christmas tallies announced U.S. job cuts every month and records the reason employers give. This is their “artificial intelligence” figure. It counts announcements and stated reasons — it does not measure what AI actually did.
- Meets the bar
- At the strictest setting — the number we publish — a layoff counts only when the company itself, in an SEC filing, a public memo, or an official statement, named AI as a reason, and the headcount is documented. Loosening the bar adds our inferences, then the headlines, so you can watch the number inflate. Even the strictest figure is a claim; it is simply the strongest one available.
- Unverified
- Blamed minus what meets the bar. Not “false”: it is blame nobody has put their name to. Some of it is real and undocumented; some is ordinary cuts dressed up as AI; some is headlines. We can't tell which from here — so we don't pretend to.
- All layoffs
- The Bureau of Labor Statistics' JOLTS survey: layoffs and discharges, seasonally adjusted. It counts how many jobs end each month. It never says why.
What this can't tell you
- Our ledger is hand-researched, not exhaustive. It covers major, named layoffs and skews toward tech and large firms. Part of the gap is coverage we haven't reached yet.
- Two different yardsticks. Challenger counts announcements; JOLTS counts separations. We show them side by side for scale and never add them together.
- A company saying “AI” is still a claim. Firms have reasons to say it — and reasons not to. We record what they said, not what is true.
- The layoffs are real either way. Whatever the cause, every number on this page is people.