AI told her Dubai Chewy Cookies were still trending — they weren't

A, a small-café owner in her 40s, was struggling to come up with a dessert side menu her customers would enjoy. She turned to AI for help — and the AI pointed to Dubai-themed products, recommending the Dubai Chewy Cookie as a still-popular dessert option.

But A hesitated. "I've heard a lot of stories about the Dubai Chewy Cookie being well past its peak, with unsold stock piling up," she said. "I wasn't sure whether to add it to the menu now. The AI says consumer awareness is still high and recommends it, but I'm not sure whether to just go along with that answer."

AI is becoming a must-have tool for small-business owners looking to run their shops more effectively. More and more are using it for menu development, recommendations and marketing — but cases like A's highlight growing concern about errors in AI responses.

Baemin Academy has drawn attention for offering guidance on how to prevent AI hallucinations and use AI more reliably in the restaurant business.

AI hallucination refers to the phenomenon in which an AI generates plausible-sounding but false information, presenting nonexistent facts as if they were real.

Baemin Academy said such errors tend to occur when users ask vague questions or give the AI multiple tasks at once.

To reduce the risk, the academy advised that the most important step is to phrase questions as specifically as possible. A representative example would be asking something like, "Tell me about trends in the Korean restaurant industry in the second quarter of this year," with precise keywords included.

Specifying the source of the information you want is equally important, the academy added. For instance, users could ask: "Based on the Baemin Restaurant Plaza website, tell me about restaurant industry trends in the second quarter of this year."

Narrowing the scope of a question is another way to improve the reliability of AI responses, the academy said. Strategies include limiting the number of answers — such as asking for "only three overseas desserts that were popular in South Korea in 2026" — or restricting results to a specific region.

The academy also advised following up with additional questions to check the consistency of AI responses. If an AI suggests three trending desserts, users can ask, "Tell me why each of these three desserts became popular," and see whether the answers hold together.

When used with a critical eye, AI can actively support every aspect of running a restaurant — from menu recommendations and development to marketing, Baemin Academy said.

Among the real-world cases the academy has highlighted, many small-business owners have used AI across a wide range of tasks — from naming their shops to building lunch menus — to run their operations more efficiently.

One notable example is B, who runs an omurice restaurant. He fed the AI information about his signature dish and the atmosphere he wanted, and used it to come up with a name for his shop.

C, who runs a small hansik restaurant, used AI to develop a lunch menu targeting office workers. The AI suggested a full meal set — stir-fried pork with three banchan — along with pricing guidance, which she used to finalize her menu.

Other cases highlighted by the academy included owners using AI to draft part-time job listings, write replies to customer reviews and create social media promotional posts.

"If you use AI well, it can work like an all-purpose new hire," Baemin Academy said. "Think of AI not as a tool but as a partner — share the situation and problems you're facing, ask for clear deliverables, and you'll find answers that take the guesswork out of running your business."