California’s AI-Unemployment Tracker offers a closer look at workers in occupations exposed to artificial intelligence. New state research raises another question: How much can job-loss predictions really tell us?
Imagine finding out that your job is on a list of occupations most likely to be replaced by artificial intelligence.
Would you start looking for another job? Sign up for classes? Spend money you had set aside for your family?
For workers already worried about paying rent, keeping up with bills or finding steady employment, a warning like that can be unsettling.
But does being on an AI job-risk list mean your job is actually disappearing?
California now has a tool to help investigate that question.
The state’s AI-Unemployment Tracker, launched in June and updated with September data on October 15, monitors unemployment claims among workers whose previous jobs are considered highly exposed to artificial intelligence.
Its latest findings offer a more complicated picture than warnings about AI taking jobs.
And new research from California’s Employment Development Department (EDD) raises another question: How much should workers trust predictions about which occupations are most likely to disappear?
What California’s AI-Unemployment Tracker actually shows
The tracker, developed by EDD and the University of California’s California Policy Lab, provides an early look at whether occupations highly exposed to AI are experiencing changes in unemployment claims.
The September 2026 update reported two notable findings:
- The three-month average of initial claims from occupations classified as having high potential AI exposure fell 4.4%, from approximately 52,200 to 49,900.
- Using a different measure based on observed AI use, the average fell 9.1%, from approximately 71,800 to 65,300.
Those averages are not seasonally adjusted. EDD also reported a modest monthly increase in its seasonally adjusted measure, within recent historical fluctuations.
So what does that mean?
The latest figures do not show a clear, accelerating wave of unemployment claims among workers in highly AI-exposed occupations.
But they don’t prove AI isn’t costing people jobs.
The tracker identifies what kind of work someone did before filing for unemployment. It doesn’t establish why that person lost the job.
A company might reduce staff because sales are down, business costs have increased, work is being reorganized or technology has replaced certain tasks.
The numbers alone cannot tell us which explanation applies.
California’s broader job market also offers a mixed picture. EDD reported that employers eliminated a net 9,300 payroll jobs in September, while the unemployment rate remained at 5.1%.
As Parriva previously reported on California’s August employment figures, statewide job totals don’t always tell the same story as the number of people actually working.
And neither measure tells us how many positions AI eliminated.
On October 15, EDD also released a research brief examining AI-exposure measures and how well an older measure of automation risk helped explain changes in employment.
Researchers examined an automation-potential measure dating from around 2016, comparing it with federal employment projections and subsequent changes in the share of workers employed in different occupations.
Their finding: The older automation measure added little explanatory value beyond existing federal employment projections.
Put simply, knowing that a job was considered vulnerable to automation did not add much to what conventional employment projections already explained about subsequent occupational changes.
That raises a question about the job-risk rankings people encounter online.
Are those rankings telling workers which jobs will disappear, or simply which tasks technology might be able to perform?
There’s an important difference.
The EDD study examined an older automation measure. It did not test every prediction involving today’s generative AI tools, including ChatGPT.
It would be a mistake to conclude that modern AI forecasts are wrong. But it would also be a mistake to treat every forecast as a reliable prediction of who will lose a paycheck.
What about Latino workers and families?
For a Latino parent working in customer service, a young person entering an office job or someone considering a career change, this debate isn’t just about technology.
It’s about whether the work they depend on will still be there.
Parriva has previously examined how AI is changing entry-level jobs in California, including research identifying automation risks in occupations employing Latino workers.
But exposure to automation is not the same as documented job displacement.
California’s tracker includes information by race, ethnicity, education and region. Still, the evidence reviewed does not establish that Latino workers in Los Angeles are experiencing a disproportionate wave of AI-caused unemployment.
There are also workers the tracker misses, including many self-employed people and those who don’t qualify for unemployment benefits.
That leaves an important part of the workforce outside the picture.
And losing a job isn’t the only way AI can affect someone’s livelihood.
As Parriva recently reported on AI-powered workplace monitoring, technology can also change how employees are evaluated, supervised and managed.
A worker might keep the same job while facing very different expectations.
Before changing careers because of AI, check the evidence
If your occupation appears on a list of jobs threatened by AI, don’t ignore it. But don’t make an expensive career decision based on that ranking alone.
Start with a few practical questions.
Is employment in your occupation actually declining? California’s labor-market information provides employment projections, wages and occupational trends.
Are employers eliminating positions or changing the skills they require? Those are different problems, and they may call for different responses.
Can you learn new skills without taking on major expenses? California’s CalJOBS offers job-search and workforce resources.
AI may eliminate some positions, change others and create new opportunities. California’s latest research doesn’t settle which jobs will be affected most.
What the tracker offers is a way to follow actual unemployment trends instead of relying only on predictions.
Because for someone trying to support a family, the question isn’t simply whether AI can do part of their job.
It’s whether there is reliable evidence that their job is going away.
