Can AI tell if your food is ultra-processed? CMC Vellore doctor breaks down what studies found

In a world where supermarket shelves are packed with packaged snacks, instant meals, and ready-to-eat foods, understanding what we actually eat has become increasingly complicated. Nutrition experts have long warned about the health risks linked to ultra-processed foods, but identifying them isn’t always easy. Now, artificial intelligence may be stepping in to help. According to insights shared by a CMC Vellore-trained neurologist, a recent study suggests that AI tools like ChatGPT may be surprisingly good at classifying foods based on how heavily they are processed.

Dr Sudhir Kumar, a neurologist at Apollo Hospital, Hyderabad, discussed the research in a post on the social media platform X, highlighting how artificial intelligence performed when tasked with identifying ultra-processed foods.

What did the study find?

The study titled "Comparing ChatGPT and DeepSeek for ultra-processed food classification: Al models for nutritional research and dietary assessment", published in the journal Nutrition, explored whether an AI language model could classify foods according to the widely used NOVA food processing classification system. This framework groups foods based on the level of industrial processing they undergo, ranging from minimally processed items to heavily engineered ultra-processed products.


Researchers asked ChatGPT to categorise 1,168 different foods into four groups: unprocessed or minimally processed foods, processed culinary ingredients, processed foods, and ultra-processed foods, often referred to as UPFs. The classifications generated by the AI were then compared with those made by a trained nutrition researcher.

ChatGPT's performance

According to the study’s findings, the AI’s performance was remarkably strong. The model achieved an accuracy rate of about 98 per cent in identifying the correct category. Its sensitivity—the ability to correctly identify ultra-processed foods—was reported at around 94.7 per cent, while specificity, which measures how well it correctly identifies foods that are not ultra-processed, reached approximately 99 percent. Only a small number of foods were misclassified.


The results were particularly notable because the model reportedly outperformed another large language model evaluated during the research.

Why does the study matter?

The ability to classify foods accurately matters for more than academic curiosity. Dr Sudhir Kumar shared that ultra-processed foods have increasingly been linked with several major health concerns. Studies have associated high consumption of these foods with obesity, type 2 diabetes, cardiovascular disease, depression, and even a higher overall risk of mortality.

Because of these connections, researchers and public health experts are paying closer attention to how much ultra-processed food people consume. However, categorising foods manually within the NOVA system can be tedious and time-consuming, especially when analysing large dietary datasets.

This is where AI could prove useful. Tools capable of rapidly sorting foods into processing categories could support nutrition research, help analyse dietary patterns, and potentially assist in large-scale public health studies.

How does AI help?

At the same time, Dr Kumar highlighted that food classification is not always straightforward. Many foods contain complex ingredient lists or undergo multiple stages of processing, which can make them difficult to categorise even for experts. Disagreements between nutrition professionals about how certain foods should be classified are not uncommon.

Because of this complexity, artificial intelligence should be treated as a supportive tool rather than a replacement for human expertise in nutrition science. Instead of making final decisions on its own, AI systems could help researchers handle large datasets more efficiently while experts provide oversight and context.

The study’s findings suggest that tools like ChatGPT may soon play a meaningful role in dietary research, offering a fast and surprisingly reliable way to analyze the processing level of foods in modern diets.