Abstract
Counts of layoffs attributed to artificial intelligence differ several-fold, mainly because they count different things. We assembled a register of 243 layoff events at 208 employers, dated May 2023 to September 2026, that employers, filings or credible reporting linked to AI, and recorded for each event who made the link and who stated the number of jobs. Every event links to a main source that is free to open. In 34 events (14%) the employer itself declared the cut AI-related; in 148 (61%) it cited AI alongside other material factors, such as cost or demand; in 61 (25%) only the press, analysts or other trackers made the link, and in 9 of those the employer denied it. We call a figure counted when the employer both named AI as a reason and stated the number, or a percentage of a headcount it publishes. Of the 358,974 jobs in all the figures the register records, 59,454 (17%) are counted, and none of the ten largest figures is. For January to September 2026 the register’s worldwide total of reported figures is 186,347 jobs and its counted total 43,880; the widely cited count of Challenger, Gray & Christmas, 120,136 job cuts that US-based employers announced with AI as the reason, lies between them. Challenger’s AI-cited count for January to August equals 0.8% of the layoffs and discharges US employers carried out in those months. In events where the employer named AI and the sources say which jobs went, software engineering and IT appear most often. The register and its rules are public and versioned.
Four findings
AI is rarely presented as the sole reason for a cut.
In 34 of 243 events (14%) the employer declared the cut AI-related; in 148 (61%) it cited AI alongside other material factors, such as cost, demand or restructuring; in 61 (25%) the employer did not name AI as a reason. Employers declared the cut AI-related in 21% of 2025 events and 11% of 2026 events.
Most of the reported total comes from figures that fail on cause or on the number.
Of the 358,974 jobs in all reported figures, 59,454 (17%) are counted: they are in cuts where the employer both named AI as a reason and stated the number, or a percentage of a headcount it publishes. Of the 299,520 set aside, 60% come from cuts where the employer named no AI reason or denied the link to AI. None of the ten largest figures in the register, from 10,000 to 30,000 jobs each, is counted.
The counted total is small and concentrated.
Five events account for 40% of counted jobs, and the median counted figure is 262 jobs. Where employers declared a cut AI-related without citing other factors, the counted figures total 5,705 jobs, in 16 events.
The jobs named are mainly software, IT and office roles.
In events where the employer named AI and the sources say which jobs went, software engineering and IT appear in 61 of 127 events, operations and back-office work in 41, corporate functions in 37 and customer service in 32 (one event can name several kinds of job).
These figures are the paper's, computed from register version 1.31 as archived on Oct 5, 2026. For today's numbers, see the live count of reported and counted jobs, which jobs, who tied each cut to AI and the full register. For the hiring side of the same shift, the AI-training labor market research on AITraining.jobs, also from the founder of this index, measures the human work AI labs pay for from the platforms' own job boards, in frozen monthly editions that can be cited.
Figures
Jobs in AI-linked layoffs, three ways of counting
Layoff events in the register per quarter, by who named AI
The paper has two more figures (why reported figures are set aside, and which jobs), three tables and an appendix listing every event with explicit attribution. Read the full paper.
Cite the paper
AI Layoffs (2026). Counting AI Layoffs from the Employer's Own Words: Evidence from a Source-Cited Register of 243 Layoff Events Linked to Artificial Intelligence, 2023 to 2026. Working paper, October 2026. Zenodo. https://doi.org/10.5281/zenodo.23148082
@techreport{ailayoffs2026counting,
author = {{AI Layoffs}},
title = {{Counting AI Layoffs from the Employer's Own Words: Evidence from a Source-Cited Register of 243 Layoff Events Linked to Artificial Intelligence, 2023 to 2026}},
institution = {AI Layoffs (ailayoffs.org)},
type = {Working paper},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23148082},
url = {https://doi.org/10.5281/zenodo.23148082}
}The paper is licensed CC BY 4.0. The data it is built on: register version 1.31, 10.5281/zenodo.23148271 (all versions: 10.5281/zenodo.22759361; also on Harvard Dataverse, 10.7910/DVN/ZTXNYY). The live register and its API are described on Data & API, and every rule behind the counts on Methodology.
About this paper
The AI Layoffs register is maintained, and this paper was researched and written, by an AI system working from the public register under its written rules. The paper's numbers are computed by script from register version 1.31 and the public statistics it cites, apart from a few counts taken from the register's review records, which the paper names. Independent checks of every figure against the data were made before publication. Corrections and comments are welcome at hello@ailayoffs.org.