How I Made $5000 in the Stock Market

AI-Exposed Jobs See Bigger Wage Gains, Research Finds

Oct 23, 2025 12:24:00 -0400 by Megan Leonhardt | #AI

Recent research shows that artificial intelligence hasn’t led to significant layoffs so far. (Dreamstime)

Key Points

Americans’ increasing use of artificial intelligence isn’t leading to major job losses just yet, but early research finds that sectors with heavy usage have experienced some wage gains.

Workers’ use of generative AI tools like large language models—such as OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude—has risen to 36.7% as of September 2025, from 30.1% in December 2024, according to new academic research published Monday from a team of economists at Stanford University, the World Bank, and Clemson University.

Interestingly, while AI has gained ground since the public release of ChatGPT in November 2022, usage patterns have been volatile. Researchers, using ongoing quarterly surveys of workers, found there was a considerable drop from the 45.9% of workers reporting using LLMs over the summer.

Still, even with the pullback in the latest survey responses, adoption in the U.S. continues to outpace other advanced economies. AI tools still are more commonly used by younger workers and in industries such as customer service, marketing, and information technology.

“Since we started the survey last December, there has been a pretty rapid increase,” says Jon Hartley, co-author of the research and an economist at the Hoover Institution. But he says the dip in AI usage since the summer is a bit of a mystery, noting that it could be a behavioral thing or perhaps employers themselves limiting the use somehow. Among those who use AI at work, the research found that about a third of workers report employing it every day.

The researchers also found that the new technology hasn’t led to any major shifts in the number of job openings or employment levels at this point. That’s in line with other recent studies that have largely found AI isn’t stealing jobs. The Federal Reserve Bank of New York released a report last month that said while AI adoption is on the rise among employers, retraining staff—rather than laying them off— is the most common outcome.

Only 1% of services firms reported that the implementation of AI resulted in layoffs in the past six months, and none of the manufacturers surveyed using AI reported layoffs.

Early research also seems to indicate that the use of AI isn’t helping workers become more productive so far, which could be helping to stem layoffs. A recent ADP report released last week, which is based on ongoing surveys of 30,000 people, found that workers who use AI on a regular basis are less likely to consider themselves as highly productive as those who rarely or never use it.

AI is not really like other workplace tools that have come online in recent decades, says Nela Richardson, ADP’s chief economist.

“It’s not a copy machine,” she jokes.

Instead, using LLMs is more akin to working with a new teammate who learns off of you. So even for heavy users, it’s a platform that’s constantly changing and evolving—and that takes time to adapt and process.

But the latest research from Hartley and his co-authors did find that more AI-exposed jobs experienced statistically significant wage gains since the release of ChatGPT.

“Much of the labor market effect of generative AI will hinge on whether it is a complement or substitute for existing tasks performed by workers of differing skill levels,” the research notes. “Complements would see enhanced wages along with productivity gains while labor substitutes would see declining or stagnant wages along with productivity gains.”

Hartley says that in regards to the wage gains, researchers found that a “pretty high fraction” of workers are using LLMs to find their next job, which is likely helping them find better matches and pay going forward.

But Hartley also doesn’t rule out that there could be a wage premium for those who successfully utilize AI. He points to the massive demand in Silicon Valley and elsewhere for workers who do AI research and programming. But more generally, Hartley says he expects that the most vulnerable workers are those who don’t learn how to use these tools to complement their work, like with all new technologies.

“It’s only been 2 to 3 years since the public release of ChatGPT. I think we’ll still learn a lot more and we’ll see what happens, “ Hartley says.

Write to Megan Leonhardt at megan.leonhardt@barrons.com