Is AI Reshaping Jobs Faster Than Past Tech?
The popular framing is that generative AI is churning the U.S. occupational mix faster than the internet or the PC did. It's a clean narrative. The actual data is blunter. Using the Yale Budget Lab's dissimilarity index, the post-GenAI shift is running about 1 percentage point higher than the comparable post-internet window — and most of the underlying shifts were already in motion before ChatGPT launched.
Key finding: In the six years after commercial-internet takeoff (1996–2002), about 7% of workers would have needed to change occupations for the mix to match the 1996 baseline. In the six years covering the ChatGPT launch (2019–2025), that number is about 8%. A 1-percentage-point difference. And the Yale authors explicitly note that occupational-mix shifts were well underway in 2021 — before generative AI was a mainstream product.
Occupational mix change in comparable 6-year windows
percentage pointsSix years after each tech wave began, the occupational mix had shifted by roughly the same amount. The difference between the internet era and the generative-AI era is about 1 percentage point — and the Yale Budget Lab notes that shifts in the mix were already underway in 2021, before ChatGPT launched.
Index: share of workers who would need to change occupations for the mix to match the baseline year (Duncan–Duncan dissimilarity index, 12-month moving average of monthly CPS data).
What's actually moving the labor market right now
The Yale Budget Lab's conclusion is that AI is present in the data, but it's not the dominant force. Three other dynamics are currently larger:
Slowing economy
Post-2022 hiring pullbacks concentrated in sectors with strong early-2020s over-hiring, well before AI integration matured.
Aging population
Retirement waves shift the occupational mix mechanically, especially in healthcare, skilled trades, and public-sector roles.
Declining immigration
Reduced inflow of foreign-born workers is reshaping occupational composition in agriculture, construction, and STEM pipelines.
AI adoption
Growing but still early-stage in its measured labor-market footprint. Effects are concentrated in specific cohorts (see our Age Gap insight) rather than broad occupational shifts.
Why this matters for AI job risk
Takeover Tracker exists to resist hype. Our methodology — task-level scoring grounded in observed AI usage — is designed to reject headline narratives that don't hold up under data. The Gimbel et al. analysis is a useful counterpoint to the widely-quoted “fastest labor-market shift ever” framing: at the aggregate occupational-mix level, the shift is real but modest and is not occurring in isolation from demographic and macro forces.
This doesn't contradict our Age Gap finding — AI's impact is showing up acutely in specific cohorts (early-career workers in exposed occupations, −20% since 2022), even as the aggregate mix moves only a little. Both can be true: sharp cohort-level displacement and muted occupational-mix churn. The distribution of AI's impact is as important as the magnitude.
For anyone reading “AI is rewriting the job market faster than ever” headlines: the aggregate data says not yet. For anyone early in their career in an exposed occupation: the cohort data says the pressure is already here. Both framings are in this data.
Sources
- Gimbel, Kinder, Kendall, and Lee (October 1, 2025), “Evaluating the Impact of AI on the Labor Market: Current State of Affairs,” The Budget Lab at Yale, and its companion report “Tracking the Impact of AI on the Labor Market.”
- The chart reproduces the post-internet value the authors quote directly (up to 7 pp by 2002, vs. 1996) and the post-GenAI value they report as approximately 1 pp higher than the internet era's trajectory over a comparable 6-year window.
- Dissimilarity index follows Duncan and Duncan's methodology: for each occupation, take the absolute difference in its share of total employment between the current month and a baseline month, then sum across all occupations and divide by two. The result is the share of workers who would need to change occupations for the current mix to match the baseline. The Budget Lab uses a 12-month moving average on monthly Current Population Survey data to reduce noise.