Is AI Destroying Jobs for Young People? Six Facts from Research

New research by Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen reveals how generative artificial intelligence (AI) is beginning to transform the U.S. labor market. They used data from the largest payroll services provider in the United States, covering millions of workers at tens of thousands of companies. They focused on the period from late 2022 to July 2025, when AI tools such as ChatGPT spread rapidly. The results show that the greatest impact is being felt by young workers aged 22–25 in occupations most exposed to AI, such as software development and customer support.

Decline in Employment Among Young People

First fact: Employment among young workers aged 22–25 in occupations most exposed to AI fell by 6% from late 2022 to July 2025. In contrast, older workers in the same fields, such as software developers and customer support representatives, saw employment growth of 6–9%. In less exposed occupations, such as healthcare assistant roles, employment grew across all age groups. The researchers used data from ADP, which includes monthly payroll records for millions of workers, and linked it to measures of AI exposure from studies such as Eloundou et al. (2024) and Handa et al. (2025).

This trend is also visible in other sources. For example, according to an analysis by the Brookings Institution, AI focuses on automating tasks, reducing the need for recent graduates in fields such as programming. Other studies show that AI is more likely to replace the “book knowledge” of younger workers than the experience of older workers.

Employment Growth, but Stagnation Among Young People

Second fact: Overall employment in the United States continues to grow, but growth among young workers aged 22–25 has stagnated since late 2022. In less exposed occupations, young people are experiencing growth similar to that of their older colleagues, but in occupations with high AI exposure, their employment has fallen by 6%. Older workers aged 35–49, by contrast, have seen growth of more than 9%. This suggests that declines in AI-exposed fields are holding back overall progress for young people.

Online sources, such as an article from The Indian Express, emphasize that this decline is particularly evident in entry-level positions, where companies are hiring fewer newcomers because AI tools automate simple tasks.

Automation Versus Augmentation: Differences in AI’s Impact

Third fact: The decline in employment is concentrated in occupations where AI automates work rather than augments it. According to the measure from Handa et al. (2025), which analyzes queries to the Claude model, employment among young people is declining in fields with high levels of automation, such as customer support. Conversely, employment is growing in occupations where AI helps augment human work, such as management roles. In fields such as software development, the number of entry-level jobs has fallen by 13%, while it remains stable in less automated roles.

Impact After Controlling for Company-Specific Factors

Fourth fact: Even after accounting for company-specific shocks, such as changes in interest rates, the relative employment of young people in the most exposed occupations has fallen by 12–13%. The researchers used regression analysis with company and time fixed effects, confirming that the trends are not caused solely by external factors. For older age groups, the effect is much smaller and statistically insignificant. This means that AI is genuinely affecting hiring, not merely reflecting broader changes within companies.

Adjustments Through Employment, Not Wages

Fifth fact: The changes are appearing in employment rather than wages. Wages for young people in AI-exposed occupations have not fallen significantly, suggesting wage rigidity. Instead of reducing salaries, companies are hiring fewer new workers.

Online analyses, such as one by Derek Thompson on Substack, add that this can be seen in data showing a decline in entry-level job postings, while wages remain stable.

Results Across Different Samples

Sixth fact: These trends persist even after excluding occupations associated with remote work or the IT sector. The researchers tested data going back to 2018 and found that AI exposure did not predict declines before 2022, including during the COVID-19 pandemic. This holds true for fields with both high and low proportions of college graduates, ruling out the impact of poorer education during the pandemic.