Frequently Asked Questions
Straight answers to the questions people actually ask, with a source for every number.
What is this site?
A public tracker of what the data shows about AI and work: where jobs are changing, how fast, and which careers and places are most affected. It aggregates a range of public data sources into two composite measures — labor-market stress and structural AI exposure. It is maintained independently by Dan Rohan; About This Site explains who runs it and why.
Sources: Data Sources · About This Site.
Which jobs are most exposed?
The ones done mostly at a computer: writing, coding, analysis, customer support. Computer user support specialists, the help desk, top our ranking, and about of the AI use there does the task outright instead of helping a person. Physical and face-to-face work shows far less.
Sources: the ranking · Anthropic Economic Index, June 2026 · BLS OEWS.
Is AI already taking jobs?
Yes, though so far in specific places rather than across the whole economy. Employers named AI as a reason for 120,136 announced job cuts through September, about 1 in 5 of all cuts this year. Employment for 22- to 25-year-olds in the most AI-exposed jobs is about 19% below where it would be if it had kept pace with peers in other jobs. The overall job market is still healthy: unemployment is around 4% and the layoff rate is low. The damage is concentrated, not widespread.
Sources: Challenger, Gray & Christmas, September 2026 · Stanford Digital Economy Lab, August 2026 · BLS, August 2026.
Who is getting hit first?
Young people trying to get their first job in an exposed field. Employment for 22- to 25-year-olds in the most AI-exposed jobs is about 19% below where it would be if it had kept pace with peers in other jobs, and the gap is still widening. Companies are mostly hiring fewer of them, not firing them. Experienced workers in the same jobs show no comparable drop. Recent college graduates also face unemployment 1.64 points above the overall rate, the widest gap in data going back to 1990.
Sources: Stanford Digital Economy Lab, August 2026 · Federal Reserve Bank of New York, 2026-Q2 · the early signals.
Are businesses actually using AI?
Yes, and more every month. 23.8% of U.S. businesses said they used AI in the past two weeks in early September, up from 17.3% last November. About a third of U.S. workers (34.6%) are in the jobs where AI does the most work.
Sources: U.S. Census Bureau BTOS, September 2026 · Anthropic Economic Index, June 2026.
Are layoffs being attributed to AI?
Yes, about 1 in 5 this year. Employers named AI as a reason for 120,136 announced cuts through September, more than any other reason. The monthly share swings a lot, from 40% in May to 9% in September. Companies choose the reason they give, so some of these are cuts they might have made anyway.
Sources: Challenger, Gray & Christmas, September 2026.
Is AI replacing or assisting workers?
Both, about evenly. In Anthropic's latest data, 53% of AI use at work helps a person and 47% does the task outright. The split matters: so far, the jobs where AI mostly automates are the ones hiring fewer young people.
Sources: Anthropic Economic Index, June 2026.
How is my area doing?
Look up a county, city or ZIP code. You get two numbers: how much of the local economy is the kind of work AI already does, and how tough the local job market is right now. Use the first to see how much your area has riding on AI, and the second to see whether trouble is already showing up. A high exposure score means a job or place has more riding on how AI plays out.
Sources: Community · Occupations.
How will I know if it's getting worse?
Watch three numbers. Young workers in exposed jobs losing ground faster than 6% a year (now 3.6%). The recent-grad unemployment gap passing 2.5 points (now 1.64). AI-cited layoffs topping 40% of announced cuts in a month (September was 9%). National Pulse updates all three monthly.
Sources: the early signals · Stanford Canaries study · NY Fed · Challenger, Gray & Christmas.
Can this site predict what happens next?
Not yet. We track eight early indicators, each with a tripwire that would mean a faster shift. One has tripped so far: job postings are holding up, which points to a gradual transition. When we can test forecasts against what actually happened, we'll publish them and grade them in public.
Sources: Indeed Hiring Lab, September 2026 · Forecast Track Record.
When you want the depth behind any number, How scores are computed has one section per score, including what each one cannot tell you. And if something looks wrong, the corrections policy explains how errors get fixed in the open.