Imagine an AI that could tell you exactly which foods are healthy for *you*, considering your unique health conditions and dietary needs. That's the promise of personalized nutritional health reasoning. But existing AI models, even powerful large language models (LLMs), struggle with the complex interplay of medical conditions, dietary habits, and food nutrition. They often fall back on generic advice or hallucinate incorrect information. A new research benchmark called NGQA aims to change this. Researchers have created a dataset that combines real medical and nutrition data from the National Health and Nutrition Examination Survey (NHANES) and the Food and Nutrient Database for Dietary Studies (FNDDS). NGQA presents AI models with challenging questions like, "Is *this specific food* healthy for *this specific user*?" The benchmark goes beyond simple yes/no answers. It also tests whether AIs can explain *why* a food is healthy or unhealthy, listing the specific nutrients that matter for that individual. For example, for a user with high blood pressure, the AI might explain that a particular food is unhealthy because it's “high in sodium.” This approach is crucial for building trust and helping people understand the reasoning behind the AI’s advice. Early tests with NGQA show that current models have a lot of room for improvement. They’re often too cautious, hesitant to give a “yes” answer without overwhelming evidence. And when they do offer explanations, they sometimes hallucinate incorrect facts or miss key contextual information. For example, an AI might flag a food as unhealthy due to high cholesterol but overlook that it's also high in protein, which is beneficial for a user recovering from opioid misuse. NGQA is a significant step toward truly personalized nutrition advice. By providing a rigorous testing ground for AI models, it pushes the field to develop more accurate, nuanced, and trustworthy AI nutritionists. However, the benchmark itself has limitations. Currently, it only covers a few health conditions, and the way it handles complex nutritional trade-offs is simplified. Future work will expand the scope of NGQA to include more diverse health conditions, social factors like food insecurity, and more complex reasoning tasks. This will pave the way for AI that can offer truly personalized guidance, helping individuals make informed dietary choices that support their unique health journeys.