Barely a day goes by when there isn’t a story about generative AI and what it means for companies and worker productivity. For good reason: The more artificial intelligence can perform job tasks independently, the more work can get done in less time. In theory, that’s great for economic growth.
But the AI jobs obsession is obscuring another important aspect of the technology’s disruption and what it means for productivity. How are generative AI tools like ChatGPT transforming life outside of work? If the technology is also accelerating people’s ability to get things done at home, that also has significant implications for the economy and in ways that aren’t reflected in traditional statistics like gross domestic product. For example, there are economic payoffs if AI is giving people more time to search for a job or learn a new skill. There are also economic benefits when there’s more time to hang out with friends or watch a movie; though hard to measure, “leisure” activities are considered by economists to have intrinsic value.
A new study by Michael Blank, a faculty fellow at the Stanford Institute for Economic Policy Research (SIEPR) and assistant professor finance at Stanford Graduate School of Business, is one of the first to examine this largely overlooked aspect of the AI revolution. Blank and his coauthors find strong evidence to suggest that the home is where AI so far is delivering the biggest productivity gains.
The researchers use internet browsing data on more than 200,000 U.S. households to track ChatGPT adoption between 2021 and 2024. They show that using GenAI helps people to perform 76 percent to 176 percent more efficiently when working on “productive” digital tasks, which the study authors define broadly as any online activity that isn’t undertaken purely for enjoyment. In other words, tasks that people normally think of as chores: for example, job hunting, travel planning or shopping for laundry detergent and other basic necessities.
By doing these productive tasks more efficiently, ChatGPT users had more free time, the researchers find. What did they do with it? They spent even more time posting on Instagram, watching Netflix or relaxing with friends.
“You can look at our findings in one of two ways,” Blank says. “The glass-half-full view is that GenAI helps people be more efficient at tedious home tasks. The glass-half-empty view is that they don’t seem to be spending their additional free time on developing new skills or other ‘human capital’ investments shown to improve standards of living.”
There is one clear downside that Blank and his coauthors — Gregor Schubert of UCLA’s Anderson School of Management and Miao Ben Zhang of USC’s Marshall School of Business — identify. The young and high earners are leveraging GenAI tools substantially faster than those who are older and lower income. What’s more, the researchers say their data indicate that the “GenAI digital divide” is widening.
That’s worrisome, Blank says, because it means that older and lower-income individuals risk losing out on the economic promise of using GenAI at home to speed up tasks like paying bills or finding answers to a health condition.
The divide is especially concerning when it comes to developing new job skills, Blank says. There’s a lot of talk about the promise of GenAI to level the playing field in education — essentially enabling anyone to have a “PhD in their pocket” — and the potential payoff for lower-skilled employees, or older workers whose jobs may be most at risk to GenAI-driven displacement. But that promise is premised on the idea that lower-income individuals are adopting ChatGPT or similar tools at home and using them, for example, to search for higher-skilled, higher-paying jobs (not binge-watching “The Pitt”).
“One of the hopes with GenAI is that these tools are going to extend what people across the socioeconomic distribution can do and especially help those at the lower end expand their capabilities and skills,” Blank says.
Blank and his coauthors say there’s an urgent need for more research into AI’s off-work impacts. Because public awareness around ChatGPT and similar GenAI tools is so high, the slow adoption by older and lower-income individuals suggests that policies are needed to help them engage with the technology — and realize its economic promise.
“Policymakers need to also consider the large share of the productivity impacts that are taking place outside of labor markets,” says Schubert, Blank’s coauthor.