How Our Health Analysis Works

The Digital Health Assistant turns a short questionnaire into a personalized nutrition report. If a tool is going to comment on your health, you deserve to know exactly how it reaches its conclusions — so this page walks through the pipeline, the data sources behind it, and just as importantly, what the assessment is not.

What you provide

The assessment asks for age, gender, height, weight and lifestyle factors — activity level, smoking, drinking — plus any everyday symptoms you choose to note. No name, no email, no account. Your responses are held under an anonymous session so the analysis can be assembled, and the results are non-identifiable.

Step 1: Body metrics, computed in your browser

The first numbers in your report — BMI (Body Mass Index), BMR (Basal Metabolic Rate), and TDEE (Total Daily Energy Expenditure) — are produced by small machine-learning models that run directly in your browser using TensorFlow. Your measurements are evaluated on your own device for these estimates. BMR estimates the energy your body uses at complete rest; TDEE scales it by your activity level to estimate your actual daily calorie needs — the anchor for every recommendation that follows.

Step 2: Your nutrient needs, from reference intakes

Nutrient targets — how much protein, fiber, vitamins and minerals a person like you needs daily — come from population reference guidelines derived from the Dietary Reference Intakes (DRIs), the nutrient standards published by the National Academies and used by U.S. health agencies. Our database carries these guidelines by age and sex, so a 7-year-old and a 70-year-old get genuinely different targets — not a scaled version of the same number.

Step 3: AI-assisted explanation

Raw numbers aren't useful without context, so the assistant uses a large language model (Google's Gemini) to organize your metrics and reference targets into readable explanations: what each nutrient does, where to find it in food and which patterns in your profile are worth watching. The model works from your computed metrics and the DRI-derived targets — it explains the analysis; it does not invent it. The same applies to the risk report: it highlights patterns associated with nutritional risk factors, ordered by significance, each with the signs and evidence behind it.

The food side: USDA FoodData Central

Nutrition data for foods on our marketplace comes from USDA FoodData Central, the federal government's food-composition database, and from producers' own measured values. Sellers' Nutrition Facts labels are generated against current Food and Drug Administration labeling regulations (21 CFR 101.9) — the same rules explained in our label-reading guide. This is what will let the marketplace match foods to household nutrition profiles when buying opens.

What this assessment is not

The Digital Health Assistant provides general wellness and nutrition information for educational purposes. It is not medical advice, a diagnosis, or a treatment recommendation, and it cannot see labs, history or anything a clinician would examine. BMI, BMR, and TDEE are population-level estimates that can miss individual context — an athlete's BMI reads “high” on muscle alone. Treat your report as an informed starting point for a conversation with a health care professional, never as a substitute for one.

Questions

We publish this methodology so you can judge the tool on its merits. If something is unclear or you believe a result is wrong, contact us — corrections make the Helper better for everyone.

Explore more guides on the Learn page, or try the complimentary health assessment.