Tech Check is a regular column on restaurant technology by Senior Editor Joe Guszkowski. It's also a newsletter.
According to its Instagram account, Ethos is the No. 1 restaurant in Austin, Texas. Specialties include a fish fry topped with caviar, a pizza hot dog, and a pastry in the likeness of Moo Deng, the famous internet hippo. The restaurant’s ingredients are locally and sustainably sourced, and it strives to create an inclusive space for customers and staff.
As far as trendy restaurants go, Ethos checks every box but one: It doesn’t actually exist. Pictures of the food, the restaurant, and even the staff are AI-generated. A chatbot on its website is unable to answer basic questions, and a reservations form leads to another website called EelSlap.com, which features a video of, you guessed it, a guy being slapped with an eel. It’s the most obvious hint to visitors that they’ve been roped in by a send-up of dining culture.
And yet that hasn’t stopped Ethos from amassing a large audience. More than 72,000 people follow its Instagram, some of whom seem to believe it’s a real place, judging by the many enthusiastic comments.
The emergence of the online restaurant business has already forced consumers to grapple with some odd questions. Is there really an establishment near me called Burger Slut? Or is it just one of the many virtual brands that now coexist with legitimate brick-and-mortars on third-party delivery apps? And if that’s the case, where is the food really coming from?
Add artificial intelligence into that uneasy mix, and it appears we now have to verify whether a restaurant exists at all.
Ethos is almost certainly a parody, and at the end of the day, it's not really hurting anyone. But it has fooled people, which is a reminder that, as restaurants forge ahead with AI, they must take care to maintain a level of trust with customers.
Indeed, restaurants are already using AI in ways that could have harmful effects on some of the trade’s unspoken agreements, such as that the food depicted on the menu is real.
As Ethos demonstrates, AI can now generate images of food with a simple prompt. This has practical applications for smaller operators who want to put pictures on their online menu but don’t have the time or budget for professional photography. More than one restaurant tech startup has offered AI image generation as a service: Type a few prompts into our interface and, voila, you have a mouthwatering burger to put on your menu.
The problem with this is obvious: Customers see one thing online and get another when the food arrives. It’s not exactly a winning formula for attracting repeat business. And, worse, it creates a sort of “boy who cried wolf” scenario: If the pictures at Restaurant A aren’t real, how can I trust the ones at Restaurant B? It’s a question consumers shouldn’t have to ask about something they plan on putting in their body.
Image generation is just one of the many ways restaurants are using AI these days. The technology is also being relied upon to whip up marketing and menu copy, answer the phone and take orders at the drive-thru. Employees are also being asked to trust it to handle things like forecasting sales and predicting inventory needs.
Like virtual brands did four years ago, this seeping-in of AI raises new, unforeseen questions for consumers. Did someone at Burger Slut really write me this happy birthday email, or was it a robot? And if it was the latter, why should it mean anything to me? Like most content coughed up by AI, it’s just soulless, hollow junk—the sort of stuff any serious restaurant should avoid like the plague.
There’s also the fact that much of this AI content is being generated by the same handful of large language models, with the eventual outcome being a sea of restaurants that all look and sound the same. The best-case scenario is that the ones who continue to do things by hand stick out like a rose in a bed of weeds. On the other hand, consumers could lose their trust in restaurants more broadly—a waning tide sinking all boats.
To be clear, I’m not opposed to AI in restaurants, because there appear to be enough benefits to outweigh the risks. But I do feel that for every piece of AI a restaurant adds, there should be a corresponding effort to bolster trust. That could be as simple as some basic transparency: A watermark on an image that indicates it was AI-generated may not help its appeal, but at least it’s honest.
Restaurants may as well get out ahead of this stuff, because the government will probably require it before long anyway. In August, the European Union enacted the Artificial Intelligence Act, a first-of-its-kind set of AI regulations that calls for AI-generated content to be labeled as such. The U.S. has created a similar framework, the AI Bill of Rights, that offers guidelines but is not legally binding.
Now, if you'll excuse me, I have to go try to make a reservation at Ethos. I'm told they're impossible to get.