All products are independently selected by our editors. If you buy something, we may earn an affiliate commission.
The most exciting food across the country today combines elements from all sorts of cuisines. The problem: what to call it, and why a name matters. “New American” has become a de facto term, but what does it mean now? Welcome to The New American Problem, a mini series exploring how we talk about contemporary American food.
The photographer Bobby Doherty used the AI-image generator Midjourney to create the images featured in this article. Each image depicts a dish that ChatGPT has identified as “New American.”
If AI were to open a New American restaurant this year, it would be called “Harvest & Hearth.” The menu would feature fried chicken and waffles, burrata, Mediterranean flatbreads, sushi burritos, hanger steak, and chocolate lava cake. Chef Thomas Keller, of landmark Napa Valley restaurant The French Laundry, or Yotam Ottolenghi, who I guess would get a stipend to move to America, would run the kitchen. Guests would dine in a setting “inspired by the natural beauty of the region,” said region unspecified, and enjoy a “relaxed, casual vibe” while listening to music by artists such as Avicii, Bruce Springsteen, Justin Timberlake, and the Black Eyed Peas.
This bonkers yet boring fictional restaurant came out of a recent conversation I had with ChatGPT, an OpenAI bot that can algorithmically process and generate text responses. Since the New American term originated over four decades ago, food writers and platforms like Yelp and Google are still slapping the label on any restaurant menu around the country that offers a glimmer of global influence, despite the label being famously murky. It’s become a meaningless, perplexing catchall for menus and chefs that don’t fit into a clearly defined (French– or Korean– or Insert Cuisine Here–shaped) box.
So I’d hoped that ChatGPT—the machine’s powerful learning skills and prowess for the written word are impressive—could help clarify what New American food actually is. (Honestly, take my job!) But after literal hours of back-and-forth, it’s clear the program is just as tired and confused about the term as those of its sources: us, the humans. In fact, ChatGPT has emphasized just how same-same food writing has been over the years. If we give the bot bad data, it’ll spit out bad answers.
Harvest & Hearth’s AI-generated “New American” restaurant playlist. Groove to its chaotic beats while you read.
I started simple. When I asked, “Explain what ‘New American’ food is,” ChatGPT typed back a paragraph much like something you’d read in a high school history class—generally hitting the right notes, but in a dry and vague way. The cuisine “draws on inspiration from diverse culinary traditions, including those of various immigrant communities in the United States,” it said. The bot offered a couple of examples of the “creative combinations of flavors and textures” one could expect within the category: “fusion dishes” such as a sushi roll with “Southwest flavors” and a Korean-style BBQ burger.
When I asked AI to generate a list of New American meals, the results were also bland and, at best, vaguely derivative of the “immigrant communities” the label is supposed to describe. Grilled salmon with quinoa, featuring roasted vegetables for a “colorful, seasonal touch.” Shrimp and grits and beef tacos were thrown in for good measure. And, out of the blue, the ramen burger, which “takes the classic Japanese noodle soup” and turns it into an all-American beef patty (in reality, the noodles become the bun).
I ventured deeper, requesting some sample New American restaurant menus, and the results were pretty forgettable. My first query yielded smoked trout dip, grilled lamb chops or mushroom risotto, and a chocolate and salted caramel tart for dessert. Another came up with an heirloom tomato salad starter followed by pan-seared scallops and stone fruit crumble. For beverages, imaginary diners could choose from sparkling lavender lemonade or ginger and honey iced tea. Not bad, but not necessarily drawn from “diverse culinary traditions.”
The fault lines really started to appear when I asked the AI for a handful of popular New American restaurants in the US. It came up with a list of objectively posh restaurants: The French Laundry, Eleven Madison Park in New York City, Alinea and The Publican in Chicago, and The Barn at Blackberry Farm in Tennessee. The label generated results more synonymous with “expensive” than diverse, the latter of which it had originally cited as a hallmark of the cuisine.
Then I prompted ChatGPT to come up with 10 influential chefs within the category. The list of exclusively white men included Dan Barber, co-owner of New York’s Blue Hill, Daniel Humm of Eleven Madison Park fame, and Stuart Brioza, the chef and co-owner of San Francisco’s The Progress and State Bird Provisions. The bot justified its choices: They’ve all “received high ratings and accolades from critics and diners.” When I widened my ask to include food influencers, ChatGPT suggested five more people, three of them Brits: Jamie Oliver, Nigella Lawson, and Ottolenghi.
Who gets to cook what—and how to describe food that straddles borders in a way that holds all of its heritage—are age-old questions. We’ve all seen the way some white chefs have handled non-European cuisines. Even AI knows: New American food can snatch “elements of other cultural cuisines, such as Mexican or Asian cuisine, without giving proper credit or acknowledgement to those cultures,” it responded, when I asked, “What are the dark sides of ‘New American’ cuisine?” Sure, it’s a canned take (also, Asian cuisine?), but it’s not entirely wrong.
In fact, none of its takes were necessarily right or wrong (except the mystifying stuff about EDM soundtracking these restaurants). That AI defines New American food as diverse, and then suggests tomato salad appetizers and sidelines chefs of color, isn’t the technology’s failing—it’s ours. The critics and gatekeepers who have been defining and unpacking the restaurant industry for decades have been majority white, and writing with a Eurocentric lens. ChatGPT is firing off contradictions because the information on which it has been trained to respond is notoriously skewed. (My former colleague and Eater reporter Bettina Makalintal came to similar conclusions when analyzing how the bot responded to diaspora food writing tropes, such as the “stinky lunchbox moment.”)
To be clear, “fusion” food, “chaos cooking,” whatever you want to call it, can be so good. Dishes like tandoori spaghetti at Pijja Palace and the Philly masala cheesesteak at Dhuaan BBQ Company in Chicago take bits of various cultures and smartly smash them together. This kind of food is understandably so beloved right now, and it’s the result of chefs cooking as much from their life experiences as they are their identities. What you’ll find in American restaurants these days is more experimental, more bold, more irreverent than ever, and, as ChatGPT is showing us, the conversation and language we use around these foods needs to catch up.
Frankly, this little ChatGPT experiment mostly produced a hazy definition and a hodgepodge of contradictions. It’s really the holes left by the chefs and the dishes not present in its responses that were most telling. Though maybe there’s a silver lining in there somewhere. ChatGPT has helped us identify certain popular restaurants, ideologies, and voices as part of the old guard; it’s drawn a line in the sand. So where do we go from here? Probably as far away from catchalls and easy-to-answer AI prompts as we can get.