When Kim Kardashian recently failed to pass the California bar exam, she blamed ChatGPT for feeding her wrong information.
You might not be daft enough to study for such an important certification using artificial intelligence shortcuts. But as a traveler, you are bombarded with unavoidable AI-generated synopses of places to eat, see, and stay.
Unfortunately, AI is transforming travel information with hallucinations that businesses are sounding the alarm about in any way they can.
Critical misinformation about dietary restrictions
London's Punjab Restaurant is one of the city's more historic places to eat. Owned by a Sikh family, the business was opened by Punjab-born Gurbachan Singh Maan in 1946. It's thought to be the oldest North Indian restaurant in the United Kingdom.
The pioneering business was among the first of its kind to stake out a spot in the then-grimy industrial district around Covent Garden, which soon became one of London's most valuable shopping districts. During the Covid-19 pandemic, the restaurant prepared more than 50,000 meals to deliver as food aid for affected people.
But Punjab's well-documented true story means nothing to AI. The restaurant's managers have had to issue urgent alerts to correct the risky nonsense being spread online.
"Don’t be so ready to Chat GPT/Gemini everything," management posted on Facebook. "A few AI tools are currently showing Punjab Covent Garden as '100% halal'—and that information is incorrect."
Some 15% of London's population observes Islam, including the city's mayor, Sadiq Khan, a former Member of Parliament. For Muslim diners, eating meat that has been prepared according to halal traditions is a fundamental dietary principle and a religious commandment.
When AI tells travelers that a certain restaurant is halal, the chatbot is sending pious followers to commit sin, according to that religion.
For a Muslim guest who might accidentally eat non-halal meat at a Sikh-owned restaurant, the transgression could have deeper ramifications given the historic tensions between the groups. If a Muslim guest felt tricked by a Sikh restaurant into eating food that's classified as sinful, it could spark a volatile incident.
Punjab did its own searches and found AI was repeating the same dangerously inflammatory lie about the business.
"Mistakes like this can be dangerous to a business," Punjab posted.
"We know this topic matters very much [to] many, especially both our Sikh and Muslim guests, and we take that seriously," the restaurant's post also stated. "While we don’t serve halal, our Muslim brothers and sisters are always welcome here at Punjab where we always have a large selection of vegetarian dishes as well as fish dishes too."
Family businesses trying to counter AI lies
While the stakes of AI-generated travel info slop are particularly high for Punjab Restaurant, business owners around the world have been urgently trying to correct the fantasies that AI is telling potential customers.
So far, programmers have not been successful at stopping AI hallucinations.
In fact, researchers have found that artificial intelligence is now lying about some subjects 35% of the time—nearly twice the rate of 18% from a year ago.
In a recent incident, Missouri pizza restaurant Stefanina's had to fend off irate customers following false info from Google AI that the pizzeria was offering deeply discounted specials that did not actually exist.
In one test, the dates of an Appalachian music festival were falsely reported; the dates AI gave were actually the dates the box office would be closed.
Sometimes AI's lies are the result of intentional sabotage.
In some cities, people are trying to poison AI to keep their favorite joints below the radar by using Reddit to post rave reviews of restaurants the posters actually believe are terrible. AI crawls the forum, steals the reviews, and then uses them to guide diners to bad restaurants.
Other complaints about the wayward tendencies of AI-generated travel information are exploding.
Promises of brunch have been made where brunch is not served.
Businesses have been described with reviews that actually pertain to another place in another country.
Foreign scam companies are creating fake images and posts on Facebook and Instagram to pretend to be family-run businesses.
User-generated reviews of hotels, shops, restaurants, and attractions are being stuffed with AI-generated fakes.
As The New York Times reported in an explosive exposé, entire travel guidebooks for sale at Amazon are being faked using AI and propped up with fraudulent 5-star reviews.
Most customer-facing AI is not actually intelligence but just a compendium of things that other people might have written about something in the past. As long as AI steals other peoples' words instead of verifying the truth on its own, the technology will continue to mislead travelers with flashes of disastrous incompetence.
The solution is simple: When you're planning travel, don't trust AI about anything—even something as simple as opening hours—and verify everything yourself.
FYI, Frommer's does not use AI to create our reviews. Every review and feature we publish is researched by a human being and checked to be correct at the time of publication.