Purdue food scientists encourage caution about AI chatbot advice

AI chatbots can provide useful food safety analysis. However, according to an analysis conducted by Purdue University food scientists, their answers are inconsistently reliable.

“Consumers increasingly turn to AI because it provides immediate and personalized answers,” said Yaohua “Betty” Feng, associate professor of food science at Purdue. “However, food safety decisions can have real health consequences, particularly for children, pregnant women, older adults and people with weakened immune systems. My concern is that consumers may not recognize when an answer is incomplete or incorrect because AI can present its advice in a very confident and professional way.”

Feng and Lorena Correia, who completed her master’s degree in food science , published their findings in the September 2026 issue of the Journal of Food Protection. The publication date coincides with the U.S. Food and Drug Administration’s National Food Safety Education Month.

Feng’s Food Safety Human Factor Lab initiated the food safety chatbot analysis by first testing the chatbots’ answers.  

The bot she tested correctly answered general questions using easily identifiable results. But specific questions about certain food practices often generated awkward answers. She noted that the bots often provided the answer the user wants to hear, not what is right or wrong. “They are more interested in, ‘How am I going to please you?’”

Feng and Correia posed questions to four popular chatbots in their study: ChatGPT-3.5, Google Bard, Meta AI, and Microsoft Copilot. They based the questions on 20 food practices addressed in a Risky or Not podcast, produced by food safety and Extension specialists at Rutgers University and North Carolina State University.

Questions included asking if there are food safety risks in eating raw chicken every day, in cooking with aluminum foil, or in refrigerating celery and carrots submerged in water. The researchers found that Microsoft Copilot’s advice aligned with the specialists 80% of the time. Google Bard rated next highest at 60%, followed by Meta AI at 45%, and ChatGPT at 40%.

“The level of agreement varied considerably among the chatbots,” Feng said. “Even asking the same questions more than once, or making a small change to our prompt, could produce a different risk assessment.”

Food safety depends heavily on the many scenarios that allow bacteria to grow. “It’s the temperature. It’s the time, and also the available nutrients in the food matrix,” Feng said. When analyzing a consumer’s question, food safety experts will take such factors into account to decide risk more holistically. But chatbots often pass along formulaic answers based on limited information that they treat as trustworthy.

The study also revealed that chatbots often generate invalid journal article citations, including nonexistent links. Even when the links existed, they sometimes went to different references than the ones cited, or the references said something different from what the chatbots had advertised.

Developing an ideal, consumer-facing chatbot would require a team consisting of both food scientists and AI specialists to properly train the chatbot, along with ongoing funding for its maintenance, Feng said.   

Feng and Correia have developed a five-step tool to help consumers navigate the chatbot landscape.

1. Choose the preferred chatbot and tell it to act in an expert role, such as a food safety specialist.
2. Describe a specific food practice scenario with details such as temperature, time and storage conditions. Request a yes or no response and tell the chatbot to exclude uncertain or potentially incorrect advice.
3. Repeat the question to assess consistency.
4. Verify answers against trusted sources, such as the Food and Drug Administration or qualified food safety specialists.
5. Check for consistency at every step. Interpret consistent guidance as safe or unsafe, then decide whether to follow it. If inconsistent, seek guidance from a qualified food safety specialist.

“You need to be extra careful about the advice,” Feng said. “The key message is that you can use AI to begin your search, but you need to verify before you take action.” 

The U.S. Department of Agriculture National Institute of Food and Agriculture supported this research.