Apart from some large swings in 2020-2021, these series visually diverge in 2024, with young workers relatively less likely to be hired into exposed occupations. Job finding rates at the less exposed occupations remain stable at 2% per month, while entry into the most exposed jobs decreases by about half a percentage point. The averaged estimate in the post-ChatGPT era is a 14% drop in the job finding rate compared to that in 2022 in the exposed occupations, although this is just barely statistically significant. (There is no such decrease for workers older than 25.)

This may provide some signal of the early effects of AI on employment, and echoes the findings from Brynjolfsson et al. But there are several alternative interpretations. The young workers who are not hired may be remaining at their existing jobs, taking different jobs, or returning to school. A further data-related caveat is that job transitions may be more vulnerable to mismeasurement in surveys.¹⁰

Discussion

This report introduces a new measure for understanding the labor market effects of AI and studies impacts on unemployment and hiring. Jobs are more exposed to AI to the extent that their tasks are theoretically feasible with LLMs and observed on our platforms in automated, work-related use cases. We find that computer programmers, customer service representatives, and financial analysts are among the most exposed. Using survey data from the US, we find no impact on unemployment rates for workers in the most exposed occupations, although there’s tentative evidence that hiring into those professions has slowed slightly for workers aged 22-25.

Our work is a first step toward cataloging the impact of AI on the labor market. We hope that the analytical steps taken in this report, especially around coverage and counterfactuals, will be easy to update as new data on employment and AI usage emerge. An established approach may help future observers separate signal from noise.

There are several improvements to be made to the present work. Our usage data will be incorporated in future updates, forming an evolving picture of task and job coverage in the economy. The Eloundou et al. metric could also be updated, to the extent that it is linked to LLM capabilities as of early 2023. And, given the suggestive results around young workers and labor market entrants, a key next step might be to look at how recent graduates with educational credentials in exposed areas are navigating the labor market.

Appendix

Available here.

Acknowledgements

Written by Maxim Massenkoff and Peter McCrory.

With acknowledgements to: Ruth Appel, Tim Belonax, Keir Bradwell, Andy Braden, Dexter Callender III, Miriam Chaum, Madison Clark, Jake Eaton, Deep Ganguli, Kunal Handa, Ryan Heller, Lara Karadogan, Jennifer Martinez, Jared Mueller, Sarah Pollack, David Saunders, Carl De Torres, Kim Withee, and Jack Clark.

We additionally thank Martha Gimbel, Anders Humlum, Evan Rose, and Nathan Wilmers for feedback on earlier versions of this report.

Citation

@online{massenkoffmccrory2026labor,
 author = {Maxim Massenkoff and Peter McCrory},
 title = {Labor market impacts of AI: A new measure and early evidence},
 date = {2026-03-05},
 year = {2026},
 url = {https://www.anthropic.com/research/labor-market-impacts},
}

References

Acemoglu, Daron and Pascual Restrepo, "Robots and Jobs: Evidence from US Labor Markets," Journal of Political Economy, 2020, 128 (6), 2188–2244.

Acemoglu, Daron, David Autor, Jonathon Hazell, and Pascual Restrepo, "Artificial intelligence and jobs: Evidence from online vacancies," Journal of Labor Economics, 2022, 40 (S1), S293–S340.

Appel, Ruth, Maxim Massenkoff, Peter McCrory, Miles McCain, Ryan Heller, Tyler Neylon, and Alex Tamkin, "Anthropic Economic Index report: economic primitives," 2026.

Autor, David H, David Dorn, and Gordon H Hanson, "The China syndrome: Local labor market effects of import competition in the United States," American Economic Review, 2013, 103 (6), 2121–2168.

Autor, David H, & Thompson, N. (2025). Expertise. NBER Working Paper, (w33941).

Blinder, Alan S et al., "How many US jobs might be offshorable?," World Economics, 2009, 10 (2), 41.

Borusyak, Kirill, Peter Hull, and Xavier Jaravel, "Quasi-experimental shift-share research designs," The Review of Economic Studies, 2022, 89 (1), 181–213.

Brynjolfsson, Erik, Bharat Chandar, and Ruyu Chen, "Canaries in the coal mine? six facts about the recent employment effects of artificial intelligence," Digital Economy, 2025.

Eckhardt, Sarah and Nathan Goldschlag, "AI and Jobs: The Final Word (Until the Next One)," Economic Innovation Group (EIG), August 2025. Available at: https://eig.org/ai-and-jobs-the-final-word/

Eloundou, Tyna, Sam Manning, Pamela Mishkin, and Daniel Rock, "Gpts are gpts: An early look at the labor market impact potential of large language models," arXiv preprint arXiv:2303.10130, 2023, 10.

Fujita, S., Moscarini, G., & Postel-Vinay, F. (2024). Measuring employer-to-employer reallocation. American Economic Journal: Macroeconomics, 16(3), 1-51.

Gans, Joshua S. and Goldfarb, Avi, "O-Ring Automation," NBER Working Paper No. 34639, December 2025. Available at SSRN: https://ssrn.com/abstract=5962594

Gimbel, Martha, Molly Kinder, Joshua Kendall, and Maddie Lee, "Evaluating the Impact of AI on the Labor Market: Current State of Affairs," Research Report, The Budget Lab at Yale, New Haven, CT October 2025. Available at: https://budgetlab.yale.edu.

Graetz, Georg and Guy Michaels, "Robots at Work," Review of Economics and Statistics, 2018, 100 (5), 753–768.

Hampole, Menaka, Dimitris Papanikolaou, Lawrence DW Schmidt, and Bryan Seegmiller, "Artificial intelligence and the labor market," Technical Report, National Bureau of Economic Research 2025.

Handa, Kunal, Alex Tamkin, Miles McCain, Saffron Huang, Esin Durmus, Sarah Heck, Jared Mueller, Jerry Hong, Stuart Ritchie, Tim Belonax, Kevin K. Troy, Dario Amodei, Jared Kaplan, Jack Clark, and Deep Ganguli, "Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations," 2025.

Hui, Xiang, Oren Reshef, and Luofeng Zhou, "The short-term effects of generative artificial intelligence on employment: Evidence from an online labor market," Organization Science, 2024, 35 (6), 1977–1989.

Johnston, Andrew and Christos Makridis, "The labor market effects of generative AI: A difference-in-differences analysis of AI exposure," Available at SSRN 5375017, 2025.

Massenkoff, Maxim, "How predictable is job destruction? Evidence from the Occupational Outlook," 2025. Working Paper.

Ozimek, Adam, "Overboard on Offshore Fears," 2019. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3777307

Tamkin, Alex and Peter McCrory, "Estimating AI productivity gains from Claude conversations," 2025.

Tomlinson, K., Jaffe, S., Wang, W., Counts, S., & Suri, S. (2025). Working with AI: measuring the applicability of generative AI to occupations. arXiv preprint arXiv:2507.07935.

1. Job offshorability: Blinder et al. (2009) and Ozimek (2019); Government growth forecasts: Massenkoff (2025); Robots: Graetz and Michaels (2018) and Acemoglu and Restrepo (2020); China shock: Autor et al. (2013) and Borusyak et al. (2022).

2. Brynjolfsson et al. (2025) compare employment trends for workers in more versus less AI-exposed occupations, using the task exposure measures from Eloundou et al. (2023) and payroll data from ADP. Johnston and Makridis (2025) do a similar task-based analysis using US administrative data, but they aggregate treatment to the industry level. Hui et al. (2024) study how freelance jobs on Upwork responded to the release of ChatGPT and advanced image generation tools, comparing workers in directly affected categories to those in unaffected categories before and after each tool's release date. Hampole et al. (2025) instrument for firm-level AI adoption using historical university hiring networks: firms that historically recruited from universities whose graduates later entered AI-related roles faced lower adoption costs.

3. Our task- and occupation-level exposure measures can readily incorporate other usage data, and be extended to different countries. We intend to apply this methodology to new settings over time.

4. In their framework, “Directly exposed'” tasks were those that could be completed in half the time with an LLM (with a 2,000-word input limit and no access to recent facts). Tasks that were “exposed with tools” were those subject to the same speedup with an LLM that had access to software for, e.g., information retrieval and image processing. Tasks that were not exposed could not have their duration reduced by 50% or more using an LLM.

5. We use the previous two Anthropic Economic Index datasets, covering usage from August and November 2025. For ONET tasks that are highly semantically similar, we split the counts across them.

6. There are judgment calls involved at every step. Should the Eloundou et al. (2023) measure enter as {0, 0.5, 1} or something else? What determines "significant" use? How do we handle tasks which seem very similar to those with high usage, but are too rare to have been picked up specifically in the sampling for the Economic Index? How much more should automation workflows count compared to augmentation? A reassuring finding which we expand on in the Appendix is that the Spearman (rank-rank) correlation of job exposure across many resolutions to these questions is exceedingly high.

7. To match O*NET-SOC codes to occ1990 codes in the CPS, we use the crosswalk provided by Eckhart and Goldschlag (2025).

8. We explore this further in three ways in the Appendix. First, we ask whether the percentile cutoff that we use to define treatment matters, varying it from the median to the 95th percentile. In all cases, the impact is flat or negative (meaning that unemployment decreases for the exposed group). Next, we focus on young workers in particular, those aged 22 to 25 as in Brynjolfsson et al. (2025). Finally, we use data on unemployment insurance claimants from the Department of Labor to measure the unemployment, rather than CPS survey responses. In no extension do we find clear impacts on exposed jobs.

9. This range is wide because the authors provide estimates against multiple counterfactuals. The 6 percentage point drop compares to a counterfactual of flat employment growth. The 16 percentage point estimate comes from a design comparing similar workers in the same firm with different occupations.

10. See Fujita, et al. (2024).