How it works

Upload the protocol (PDF, Word, ODT, RTF, TXT, Markdown or HTML), paste the criteria, or enter a ClinicalTrials.gov ID.

Files are read in your browser; only the text is sent. You see exactly what we received before anything else happens.

Criteria become rules over survey measurements: age, BMI, blood pressure, HbA1c, kidney function, diagnoses, medicines.

Five validated trials use hand-written rules; for others an AI translates, using only a fixed list of survey questions. You can mark any rule as wrong; it is left out.

The rules run on CDC NHANES. Each person stands for a known number of Americans, so the result is a national count with a range.

The AI never produces a number. Every figure comes from the engine.

Share and count of US adults with a range, the rules that exclude the most people, the one rule worth relaxing, what can't be checked, every assumption and a PDF.

The headline number is free. The full report is $349, one-time.

People are interviewed at home, then examined in a mobile clinic: height, weight, blood pressure, blood and urine tests, medicines and diagnoses.

Measured: BMI, blood pressure, HbA1c, glucose, cholesterol, kidney and liver tests. Asked: diagnoses, smoking, pregnancy, PHQ-9. Medicines: prescriptions seen by the interviewer.

A person passes if they meet every inclusion rule and no exclusion rule. The share is the weighted sum of people who pass over the weighted total.

The range comes from the survey design (strata and clusters). Missing answers fail a "must have" rule and pass a "must not have" rule.

Rules are applied one after another. We also drop each rule on its own and recount: the one that adds the most people back is your best candidate to relax.

AI maps criteria to a fixed list of 83 survey variables. It can't invent variables or run code, every rule is re-checked, and you review them before the count.