Every week seems to bring new conclusions about what AI will bring to the labor market. “AI will eliminate jobs and create new jobs. It will replace entry-level workers, increase productivity, transform professions, and make hiring more efficient.” The harder questions are more direct. What is AI actually doing to the labor market right now?
This week, Revelio Labs is launching a new monthly AI labor market tracker in an effort to answer that question using data on employee counts, job listings, salaries, and employee sentiment. The tracker tracks the impact of AI across five parts of the labor market: worker supply, employer demand, employment and wage equilibrium, the activities that make up a job, and the process by which workers and employers find each other. It combines Revelio Labs’ original analysis with modern applications, replications, and extensions of findings from the latest academic research.
This goal is intentionally descriptive rather than predictive or prescriptive. We’re not trying to predict the ultimate future of work or tell companies how they should leverage AI. We want to measure the changes that are already happening across a comprehensive set of indicators.
Various employment stories told by exposure to AI and introduction of AI
A central challenge in studying the labor market impact of AI is to distinguish between expectations and adoption. Companies may adjust hiring plans before deploying AI tools, while workers may change their skill investment decisions in response to forecasts of future demand. Conversely, measurable labor market impacts are likely to emerge only after organizations have formally adopted AI technologies. It helps to distinguish between these two ideas, which are often treated as interchangeable.
The first is AI exposure, or the proportion of work activities performed in an occupation that current AI systems can probably perform. Exposure does not mean that employers or workers in these roles are already using AI. Behavior changes just because there is a possibility. Companies may reduce hiring in high-exposure roles, shift demand to other jobs, or reconsider needed entry-level positions before fully implementing the technology. These are anticipatory effects, and changes in the labor market are driven by what employers expect AI to be able to do, rather than by the observed use of AI within companies.
Building on the approach of Brynjolfsson, Chandar, and Chen (2025) in Canaries in the Coal Mine, we find that employment growth in the most AI-exposed occupations has been approximately 4% less than that in the least AI-exposed occupations since the inception of ChatGPT. We found that the attrition was greater for younger workers in exposed occupations compared to older workers.
The second idea is AI adoption, whether companies are actually starting to use AI in their operations.
Following Hosseini Maasoum and Lichtinger (2025), we identify AI-adopted companies from job postings. A company is considered to be hiring AI when it posts a job opening for an AI integrator role. Note that this approach to identifying adopters differs from Kharazian, Simon, and Stevens (2026), who identified adopters from spending data.
Since October 2022, hiring companies have increased their headcount by 27% compared to non-hiring companies. The hiring company was growing faster than before. This means that recruitment is not random and that adopting firms are different from non-adopting firms prior to adoption. You can also see that the growth is not evenly distributed. Senior positions will increase by 31%, while junior positions will increase by only 6%.
The findings between AI exposure and adoption are not contradictory. AI reduces the demand for certain types of work while helping the overall growth of companies that use it.
Work changes faster than work
Employment numbers and job titles are only part of the impact of AI.
A job is a bundle of activities. Employers can make significant changes to these activities without changing the occupation or eliminating the job altogether. While continuing to perform the same role, employees can spend less time drafting, summarizing, data entry, and performing routine analysis and more time reviewing AI-generated output, exercising judgment, and managing AI-assisted workflows.
Our activity dissimilarity index (DI) measures how different the mix of jobs performed across the economy is from the mix observed in the previous year, weighted by the number of workers doing them. It has remained relatively stable, but has risen sharply over the past three months.
In June 2026, the year-on-year change in the activity composition dissimilarity index increased to 8.4 percentage points. This means that 8.4% of the economy-wide headcount-weighted activity mix would need to be reallocated across activities to return to the June 2025 mix.
Interestingly, most of this change is occurring within occupations, rather than because the economy is shifting from one occupation to another. Roles and job titles may remain the same even if the activities that a worker performs in the job change.
This helps to explain why studies that focus on occupation-level employment generally show relatively limited change. If jobs remain the same and the jobs within them change, an analysis based primarily on occupational distribution will miss much of the change.
There is also evidence that employers are moving away from tasks that are most easily performed by AI. Since November 2022, the proportion of job listings with high AI exposure has decreased by approximately 5 percentage points.
This decline does not prove that this job has already been replaced by AI. Employers may also be reacting to expectations of what AI will soon be able to do, as well as other changes in labor demand. However, this pattern is consistent with some automation, or reduced reliance on compromised activities, that is already taking place.
The most obvious explanation at this point may be: AI automates tasks, but it doesn’t automate jobs.
How AI makes hiring less efficient
AI will not only change what employers hire people to do; The way candidates compete for jobs is also changing.
The recruitment process has fewer successful matches. The number of job postings required for each hire has increased over the past six years, while applicants are increasingly reporting silence and lack of communication from hiring managers. These trends predate ChatGPT, so they can’t be completely blamed on generative AI. But by allowing candidates to submit more sophisticated applications, AI could exacerbate existing problems by weakening the signals employers use to identify strong candidates.
The jobs-to-jobs ratio, or the number of jobs a company needs to post to hire one employee, has been increasing for six years and will begin to increase dramatically in late 2022. Whereas previously it was around 1 or less (meaning companies hire more than one person for each post), this ratio is now over 5. To hire one employee, an employer would need to post a job an estimated five times. Employers advertise jobs, but with each job ad posted, the number of successful matches decreases.
Increasingly, applicants are reporting poor communication from their employers. Recruiter mentions of ghosting in non-offer interview reviews have increased by approximately 9% since October 2022.
Generative AI may be exacerbating existing matching problems.
Candidates can now tailor their resumes, draft cover letters, prepare responses, and submit applications at little to no additional cost. This makes it easier for strong candidates to present themselves well, but it also makes it easier for most people to craft an application that looks plausible. As a result, the signal value of a strong resume decreases.
Employers may receive more applications without receiving better signals about who will perform well. As it becomes harder to identify strong candidates, employers communicate with fewer applicants and candidates compensate by submitting more applications.
The evidence does not prove that generative AI caused a decline in the number of hires per post. This trend started long before ChatGPT. Nor have we directly observed whether AI-generated applications are causing an increase in recruiter silence. But the results suggest that AI may be adding noise to a hiring process that already struggles to produce effective matches.
Track the impact of AI on the labor market in real time
The impact of AI will not emerge as a single labor market shock. They emerge through changes in recruitment, corporate growth, career entry, job activities, wages, and the mechanisms that connect workers and employers.
These changes can go in different directions. Employment may weaken in at-risk occupations while growing in hiring companies. A job can remain the same even if the work inside it changes. While recruiting tools become more powerful, the hiring process can become less efficient.
That’s why we built the AI Labor Market Tracker. We update monthly, adding new metrics and revisiting existing results as new workforce and job data becomes available. Our goal is to provide a consistent view of how AI is currently changing the labor market.
