This paper presents a study on integrating expert knowledge into Large Language Models (LLMs) to create a customized nutrition assistant that provides accurate and personalized food product explanations. The research team, consisting of computer scientists and nutrition experts, collaborated with registered dietitians (RDs) to evaluate the strengths and weaknesses of LLMs in generating nutrition information. Through a mixed-methods approach, RDs validated GPT-4 outputs at different levels of prompt specificity, leading to the development of design guidelines for prompting LLMs. These guidelines were used to create a GPT prototype, The Food Product Nutrition Assistant, which was refined and evaluated in focus groups with RDs. The study found that using dietitian-reviewed template instructions improved the generation of detailed food product descriptions and tailored nutrition information. LLMs have the potential to contribute to nutrition and dietetics by generating food product explanations that facilitate informed food selections. However, the accuracy and reliability of these models remain unverified. The study highlights the importance of integrating expert knowledge into LLMs to ensure the safety and efficacy of food explanations. The research team developed a set of design guidelines based on feedback from RDs to enhance the performance and face validity of GPT-4 in providing food explanations. They also created a customized GPT prototype using these guidelines and assessed its effectiveness in a collaborative focus group with RDs. The study addresses the need for food explanations that improve food literacy, the role of RDs in health coaching, and the current capabilities and limitations of AI, particularly LLMs, in nutrition and dietetics. The research also explores the challenges and prospects of AI integration in nutrition and dietary guidance, emphasizing the need for rigorous validations and transparency of limitations to ensure the safety and efficacy of AI-generated food explanations. The study concludes that integrating expert knowledge into LLMs can enhance their utility in providing nutritionally sound and personalized food product explanations. The paper makes several contributions, including an empirical understanding of LLMs in generating food product explanations, the formulation of design guidelines for future research, and the development of a customized GPT prototype for food product explanations.This paper presents a study on integrating expert knowledge into Large Language Models (LLMs) to create a customized nutrition assistant that provides accurate and personalized food product explanations. The research team, consisting of computer scientists and nutrition experts, collaborated with registered dietitians (RDs) to evaluate the strengths and weaknesses of LLMs in generating nutrition information. Through a mixed-methods approach, RDs validated GPT-4 outputs at different levels of prompt specificity, leading to the development of design guidelines for prompting LLMs. These guidelines were used to create a GPT prototype, The Food Product Nutrition Assistant, which was refined and evaluated in focus groups with RDs. The study found that using dietitian-reviewed template instructions improved the generation of detailed food product descriptions and tailored nutrition information. LLMs have the potential to contribute to nutrition and dietetics by generating food product explanations that facilitate informed food selections. However, the accuracy and reliability of these models remain unverified. The study highlights the importance of integrating expert knowledge into LLMs to ensure the safety and efficacy of food explanations. The research team developed a set of design guidelines based on feedback from RDs to enhance the performance and face validity of GPT-4 in providing food explanations. They also created a customized GPT prototype using these guidelines and assessed its effectiveness in a collaborative focus group with RDs. The study addresses the need for food explanations that improve food literacy, the role of RDs in health coaching, and the current capabilities and limitations of AI, particularly LLMs, in nutrition and dietetics. The research also explores the challenges and prospects of AI integration in nutrition and dietary guidance, emphasizing the need for rigorous validations and transparency of limitations to ensure the safety and efficacy of AI-generated food explanations. The study concludes that integrating expert knowledge into LLMs can enhance their utility in providing nutritionally sound and personalized food product explanations. The paper makes several contributions, including an empirical understanding of LLMs in generating food product explanations, the formulation of design guidelines for future research, and the development of a customized GPT prototype for food product explanations.