![Pixelated illustration of a broom on a fuchsia background](https://cdn.theatlantic.com/thumbor/0sQTBC5cAo-M6owGABu6TbQZ068=/0x1:4167x2345/960x540/media/img/mt/2023/04/AI_Chores/original.jpg)
ChatGPT is revolutionizing work. The AI-powered chatbot, which can write sophisticated responses to just about any prompt and has passed an MBA exam, is holding its own as a coder and is already helping with professional writing. In the few short months since its public launch in 2022, it’s transformed the future of white-collar labor, and provoked an intense debate: Will AI steal our jobs? But during this same period, ChatGPT has also begun quietly shaping work in another, less heralded—but equally influential—realm: the home.
Just as workers have been trying out the software to see how it might make office tasks easier, others have been experimenting with how it might lighten the burdens of unpaid domestic work. Many are using it to plan meals and generate grocery lists. For some it has become an ad hoc family-trip planner and scheduling assistant. Others are testing its ability to budget for a house and make up bedtime stories.
ChatGPT and other forms of generative AI are poised to become a fixture in American homes, similar to other virtual assistants like Alexa. While chatbots could change how certain household tasks get done, it’s less clear whether they could really make a dent in our most persistent domestic challenges: the significant amount of time still spent on chores and the inequality in how that labor is divided.
Folks who have started using ChatGPT for their personal to-do lists have found it incredibly useful, especially for one-off tasks. For Raina Kumra, the founder of the company Spicewell and a mother of two in Santa Monica, California, lining up summer camps for her kids each year has been “an entire job.” Many camp options she’s interested in last only one week, so finding enough that didn’t conflict with one another—to cover all of summer vacation—involved making spreadsheets, researching options, and cobbling together a calendar. After all that, she still had to register online, coordinate carpools as needed, find child care for the afternoon if the camp was a half day, and pay all the fees for both of her kids, which could amount to thousands of dollars. This year, she was dreading it. “I turned to AI to help simply out of desperation,” she told me. Within seconds of entering her query, the software generated a comprehensive schedule that (with a few edits) she actually plans to use. It included a good range of activities, from traditional options like sports, art, and nature, to more unexpected choices, such as robotics, cooking, and yoga. This saved her dozens of hours of work, she said.
Unfortunately, there were still dozens of hours of work left to do. Although the AI found the summer camps and made the schedule, it couldn’t complete any of the other steps—including perhaps the biggest one of driving her kids to camp each day. As much as ChatGPT can do, there’s so much it can’t touch. It might share recipes, but it can’t cook meals. It can make a chore chart, but it can’t wash the dishes or take out the trash. Even the steps that it does help with still need oversight. Most people would want to scrutinize a ChatGPT-generated financial plan before implementing it, for example. And despite the clever scheduling it did for Kumra, some of the summer-camp names it gave her were slightly off —she was still able to find them online, but it took more Google sleuthing. Some people who have used the software for family-vacation planning have complained about clichéd suggestions and recommended spots being closed (perhaps because the software has “limited knowledge” of the world after 2021).
In the coming years, many of these kinks will likely be sorted out. But Ekaterina Hertog, the leader of the University of Oxford’s ongoing DomesticAI research project, which is examining what AI has done (and will do) to household labor in the U.K. and Japan, points out that more advanced AI won’t necessarily directly translate to time saved on housework. The introduction of new technology into the home shapes chores in both intentional and unintentional ways. Take washing machines, which seemed like they would easily slash the time people used to spend cleaning clothes by hand. And they did reduce the burden of laundry, but not as much as you might expect, because after their introduction, hygiene standards shot up, and people began to clean their garments more frequently.
Hertog suspects that a similar situation could play out with AI. It will likely take practice to learn to write prompts that get the most relevant and helpful responses. And if the software becomes embedded in home life, how much work will it be to teach kids to use it (and to supervise them to make sure they’re doing so safely)? Even more immediately, will having easy access to novel meal plans raise our standards for variety in our diets, leading to more time cooking and shopping?
That second scenario seems to already be playing out for Carole Alalouf, the owner of an animation company who lives in Montreal with her two teens and her husband. Because she has a family of picky eaters—she’s on a Mediterranean diet, her husband is keto, and two of her family members don’t eat pork or shellfish—she cooks the same things over and over, which “drives me bananas,” she told me. ChatGPT instantly gave her some fun new ideas. One of her favorites was Saganaki, a Greek dish featuring fried halloumi cheese, which she has enjoyed at restaurants but never thought to make at home. But though it can spare you from having to search for recipes and make a grocery list, AI-assisted meal planning risks adding some of those hours back in time spent tracking down tough-to-find ingredients or preparing unfamiliar meals more slowly.
If ChatGPT can save even an average of a few minutes a day, that would be significant progress given that on a population level, the time devoted to housework hasn’t meaningfully budged in more than two decades, Melissa Milkie, a sociologist at the University of Toronto, told me. Plus, because women do so much more housework than men—moms in heterosexual marriages do 1.7 times as much as their husbands, for example—they also stand to benefit more from any reduction. Kumra thinks “ChatGPT is 100 percent going to help moms with all our unpaid labor,” she told me over email.
But time is only one part of the gendered inequality in domestic work. According to Eve Rodsky, the author of the book Fair Play, a task has three components: The first is conception, or noticing what needs to be done; then comes planning; and then comes execution, or actually doing it. ChatGPT is great for planning and can be useful in execution too. But as of yet, it really doesn’t help with conception. “It will never be able to tell you that your second son needs his adenoids taken out. It will never be able to help with the noticing that your child’s bangs have grown out into their eyes,” Rodsky told me. Noticing and keeping track of everything that needs to be done is “where the mental load really lives.”
One common way of thinking about how domestic labor plays out with many heterosexual couples is that women tend to act as managers and men as helpers, assigned chores by their wives. ChatGPT is perhaps best seen as another helper. It will—sometimes clumsily and sometimes brilliantly—complete the tasks delegated to it, but an AI is no substitute for an equal partner. Rather, it’s merely a reflection of the culture that created it. The software’s occasionally sexist outputs are proof that it sometimes mirrors back our worst impulses. But for couples who work to create a fair distribution of labor in their household, it could also have the potential to mirror back our best intentions. The best-case scenario is for artificial intelligence to become a home helper who reports to two equal managers.