This paper explores the integration of Large Language Models (LLMs) in providing personalized nutrition information, particularly through the development of a customized GPT prototype called *The Food Product Nutrition Assistant*. The study aims to enhance the accuracy and reliability of LLM-generated food product explanations by collaborating with registered dietitians (RDs) to evaluate and refine the model's outputs. **Key Findings:** 1. **Input Specificity:** RDs found that outputs with more detailed input specifications (Level 2 and Level 3) were more comprehensive and accurate, while Level 1 inputs were less detailed and often lacked specific nutrient information. 2. **Reliability and Accuracy:** Outputs often contained irrelevant or misleading information, such as incorrect allergen claims and imprecise health claims, which could mislead consumers. 3. **Educational Value:** The outputs lacked educational context and often used technical language that might be confusing for consumers with lower reading proficiency. 4. **Customization:** While dietary goals improved the personalization of outputs, they did not always align with broader dietary needs or provide sufficient contextual information. **Design Guidelines:** 1. **Provide Product Label Information:** Include Nutrition Facts labels and ingredient lists in the prompt. 2. **Include Dietary Guidelines and MyPlate Sources:** Use credible sources like the Dietary Guidelines for Americans and MyPlate guidelines. 3. **Include Example Sentences:** Provide structured sentences to guide the model's output, especially for allergen wording and cost implications. 4. **Limit Words and Terminology:** Avoid generic terms and focus on specific nutrient amounts. 5. **Instruct for Comprehensiveness and Balanced Diets:** Ensure outputs consider the broader context of a balanced diet. 6. **Specify Functionality and Scope:** Emulate a virtual dietitian and avoid giving medical advice. 7. **Specify Guidelines on Nutrient Content:** Use standardized terms and define nutrient categories. 8. **Include Comprehensive Nutritional Guidance and Food Pairing Suggestions:** Provide detailed nutritional guidance and food pairing suggestions. **Prototype Development:** The research team developed a customized GPT prototype using these guidelines. The prototype was refined through focus groups with RDs, who provided feedback on the outputs and contributed to further improvements in the template instructions. **Conclusion:** The study demonstrates the potential of LLMs in generating detailed and personalized nutrition information, but highlights the need for careful validation and the integration of expert feedback to ensure reliability and accuracy. The customized GPT prototype, *The Food Product Nutrition Assistant*, aims to address these challenges and enhance consumer education on dietary choices.This paper explores the integration of Large Language Models (LLMs) in providing personalized nutrition information, particularly through the development of a customized GPT prototype called *The Food Product Nutrition Assistant*. The study aims to enhance the accuracy and reliability of LLM-generated food product explanations by collaborating with registered dietitians (RDs) to evaluate and refine the model's outputs. **Key Findings:** 1. **Input Specificity:** RDs found that outputs with more detailed input specifications (Level 2 and Level 3) were more comprehensive and accurate, while Level 1 inputs were less detailed and often lacked specific nutrient information. 2. **Reliability and Accuracy:** Outputs often contained irrelevant or misleading information, such as incorrect allergen claims and imprecise health claims, which could mislead consumers. 3. **Educational Value:** The outputs lacked educational context and often used technical language that might be confusing for consumers with lower reading proficiency. 4. **Customization:** While dietary goals improved the personalization of outputs, they did not always align with broader dietary needs or provide sufficient contextual information. **Design Guidelines:** 1. **Provide Product Label Information:** Include Nutrition Facts labels and ingredient lists in the prompt. 2. **Include Dietary Guidelines and MyPlate Sources:** Use credible sources like the Dietary Guidelines for Americans and MyPlate guidelines. 3. **Include Example Sentences:** Provide structured sentences to guide the model's output, especially for allergen wording and cost implications. 4. **Limit Words and Terminology:** Avoid generic terms and focus on specific nutrient amounts. 5. **Instruct for Comprehensiveness and Balanced Diets:** Ensure outputs consider the broader context of a balanced diet. 6. **Specify Functionality and Scope:** Emulate a virtual dietitian and avoid giving medical advice. 7. **Specify Guidelines on Nutrient Content:** Use standardized terms and define nutrient categories. 8. **Include Comprehensive Nutritional Guidance and Food Pairing Suggestions:** Provide detailed nutritional guidance and food pairing suggestions. **Prototype Development:** The research team developed a customized GPT prototype using these guidelines. The prototype was refined through focus groups with RDs, who provided feedback on the outputs and contributed to further improvements in the template instructions. **Conclusion:** The study demonstrates the potential of LLMs in generating detailed and personalized nutrition information, but highlights the need for careful validation and the integration of expert feedback to ensure reliability and accuracy. The customized GPT prototype, *The Food Product Nutrition Assistant*, aims to address these challenges and enhance consumer education on dietary choices.