Who’s using AI at home and how do they use it? And is it making them more efficient?
These are questions Michael Blank, an assistant professor of finance at Stanford Graduate School of Business, wanted to answer in his recent research. “Much of the research and policy focus around AI has been about labor markets,” he says. “But adoption of AI has become more rapid and persistent among households than at work. To change how people work takes changing a lot of things — like teams agreeing on a chatbot — but adopting these tools at home means just typing in a request, like finding the cheapest airline tickets.”
As tools, large language models (LLMs) could “be transformative for society,” Blank says, “enabling people to do technologically sophisticated things and learn new, marketable skills without formal education, like coding.” But achieving that kind of democratization of learning would require equal access to — and usage of — AI tools across socioeconomic groups.
To understand who’s using AI and how they’re benefiting from it, Blank and his collaborators Gregor Schubert of the University of California, Los Angeles, and Miao Ben Zhang of the University of Southern California analyzed how thousands of U.S. households use their online time. They found that younger, higher-income households use generative AI, specifically ChatGPT, to improve their productivity and use the time they’re saving for leisure activities.
“We see that people are using ChatGPT to improve from a productivity standpoint, giving them more time to spend on things they enjoy. But it’s not happening in an equal way across the income and age distributions,” says Blank, a faculty fellow at the Stanford Institute for Economic Policy Research.
The researchers analyzed the browsing history of more than 200,000 households between 2021 and 2024. Among the data points they examined were when a household first used ChatGPT and which sites it visited around the time it was using AI.
Quote
People are using ChatGPT to improve from a productivity standpoint, giving them more time to spend on things they enjoy. But it’s not happening in an equal way.
Author Name
— Michael Blank
By the end of the period studied, 16% of all households had used ChatGPT at least once. The researchers found that people between the ages of 18 and 34 and those with annual incomes over $100,000 (especially over $200,000) were much more likely to use ChatGPT. “They have the most aggressive adoption,” Blank says. By late 2024, 24% of households headed by someone 34 or younger had used ChatGPT, versus 11% of older households. Younger, wealthier households also began using ChatGPT sooner and were more likely to continue using it than other demographics.
This pattern of findings suggests a clear digital divide in GenAI adoption and usage, a gap that the authors note has widened over time.
More Time for Leisure
The researchers also looked at how people use AI and how it impacts their available time. “We were especially interested in understanding whether unequal adoption leads to unequal benefits for different types of households,” Blank says.
That meant first distinguishing between types of activities, Blank notes. “There are things we do with our computers or smartphones that economists call ‘home production,’ like paying bills or figuring out travel arrangements, that are productive but don’t result in intrinsic enjoyment. Then there are intrinsically enjoyable things like going on social media or watching movies or browsing sports information — these are leisure activities,” he explains.
Blank and his collaborators found that people were using ChatGPT for more productive tasks, as suggested by the sites they visited during, before, and after their LLM sessions: for example, IRS.gov and TurboTax during tax season, or Canvas and Blackboard when doing educational tasks. “Presumably they’re finding LLMs more efficient than traditional means,” Blank says.
The next question, then, is what people did with the time gained from using AI. Here, the researchers found that people used that time for leisure activities. Much of that activity was online, including time spent on nonproductive tasks on ChatGPT and social media sites. “ChatGPT adopters are able to use the time freed from efficiency gains there to do things they get intrinsic enjoyment out of,” Blank says.
By the numbers, the research shows that households using ChatGPT increased the proportion of time they spend on leisure activities by 31 percentage points and decreased the share of time on productive sites by 21 percentage points. While overall browsing time went up for households using ChatGPT, the time they spent on productive tasks remained unchanged while leisure time rose dramatically, suggesting they were getting more done — up to an estimated 176% improvement in efficiency.
Understanding the use of GenAI at home helps paint a fuller picture of AI-related productivity and value. “So far we haven’t seen a noticeable uptick in AI-related productivity for businesses in terms of output and revenue per unit of labor or unit of capital,” Blank says. “But the fact that people can do more enjoyable activities after adopting ChatGPT suggests a beneficial impact on productivity: People can consume more of what they want because they don’t have to spend as much time on what they see as chores.”
The challenge is that not everyone is experiencing these gains. “There’s a pretty stark gradient across income and age distributions of who’s adopting these tools at home and experiencing related quality-of-life gains,” Blank says. “If there were more equitable take-up at home, these gains would be shared more equally rather than potentially worsening inequality.”
Despite its transformative potential, AI is following an age-old story of technological adoption, Blank notes: “Throughout the history of technological improvements, it tends to be higher-income people who are able to adopt improvements most rapidly.”
