The use of the internet for accessing health information has significantly risen in recent times. However, studies have revealed that almost half of the nutrition-related content available online is inaccurate or of substandard quality. While artificial intelligence (AI) holds promise for transforming digital healthcare, its efficacy in addressing nutrition queries remains uncertain. To bridge this knowledge gap, researchers conducted a study published in JAMA to evaluate AI’s capacity to provide accurate energy and nutritional details for 222 food items across various languages.
In this study, two AI chatbots, namely ChatGPT-3.5 and ChatGPT-4.4, were assessed for their reliability in furnishing calorie and nutrient data for eight adult menus to draw precise conclusions. The study leveraged the expertise of the Japanese Food and Drug Administration to validate the accuracy of AI responses by evaluating consistency through the coefficient of variation. Statistical analyses were performed using SPSS Statistics version 26.
The findings indicated that while there were negligible discrepancies in the estimations of energy, carbohydrates, and fats between nutritionists and AI, there was a notable variance in protein estimation. Both AI systems accurately provided energy content for 35% to 48% of the 222 food items within a 10% margin. ChatGPT-4 exhibited superior overall performance, albeit with a tendency to overestimate protein content.
Experts believe that AI holds immense promise in the realms of public health and nutrition. A review published in a 2023 edition of Nutrients highlighted the effectiveness of AI models in identifying areas with limited access to nutrient-dense foods and in visualizing and assessing food environments. Furthermore, AI is seen as capable of predicting regional nutritional requirements, tracking global food trends, and standardizing dietary guidelines worldwide for consistent messaging.
The authors of the Nutrients study emphasized the vast potential of AI in public health nutrition, suggesting that current research merely scratches the surface of what is achievable. They urge fellow researchers to explore innovative applications of AI to promote healthier diets and improve population-level dietary health.
While AI chatbots are deemed valuable for collecting energy and nutrient data and offering real-time food analysis, their inability to provide personalized nutritional guidance remains a drawback. The generic nature of AI chatbots limits their capacity to deliver tailored dietary recommendations, including precise nutrition advice and portion sizes. These limitations stem from AI’s lack of specialization in nutrition and health, leading to challenges in offering appropriate household units for consumers. Future advancements are crucial to enhance the accuracy and utility of nutrition information provided by AI technologies.
